We develop theoretical and practical large scale scientific machine learning tools for understanding global-scale atmospheric processes and high-dimensional engineering turbulence.

Publications: https://scholar.google.com/citations?user=wtHkCRIAAAAJ&hl=en

Use ML methods to analyze geodetic time series and their derivatives.

NextG network optimization and control with low-latency AI: With the advent of software-defined networking, there has been a paradigm shift in the cellular network landscape, leading towards virtualization and disaggregation. Open RAN promotes open interfaces to facilitate data collection for comprehensive network analysis. Furthermore, it also promotes embedding control hooks directly into cellular network functions. These are critical steps towards actualizing the vision of an intelligent network, where optimization and control are not just possible, but are fundamental aspects of the network’s operation. We incorporate low latency AI capabilities to cellular networks to make realtime network decisions possible, this is crucial to enable the diverse range of end user applications anticipated in the next generation of mobile communication.

Compute for Alex Cloninger's Lab. Admin: Alex Cloninger

We are working on developing new technologies and cross-stack solutions to improve the performance and energy efficiency of computer systems for emerging applications.

Publications:

Temporary testing environment for k8s experimental resources

As the scale, complexity, and variety of data grows (aka Big Data), the use of machine learning (ML) and artificial intelligence (AI) techniques to make sense of, and interact with, such data — collectively called predictive data analytics, statistical data analytics, ML-based data analytics, or simply advanced data analytics (also ADA!) — is increasingly critical for data-driven applications in the enterprise, Web, science, and other domains. Alas, building and deploying ML/AI-powered data analytics applications still involves far too many bottlenecks that slow down the lifecycle of such applications, raise costs, frustrate many application users, and in some cases, make high-quality data-driven decision making almost impossible. The mission of the ADALab is to democratize advanced data analytics by making it dramatically easier, faster, and cheaper to build and deploy ML/AI-powered data analytics applications throughout their lifecycle.

Publications: CHAP-child: An open source method for estimating sit-to-stand transitions and sedentary bout patterns from hip accelerometers among children. Jordan A. Carlson et al. (15 authors). International Journal of Behavioral Nutrition and Physical Activity 2022. CHAP-Adult: A Reliable and Valid Algorithm to Classify Sitting and Measure Sitting Patterns Using Data from Hip-Worn Accelerometers in Adults Aged 35+. John Bellettiere et al. (14 authors). Journal for the Measurement of Physical Behaviour 2022. Cerebro: A Layered Data Platform for Scalable Deep Learning. Arun Kumar, Supun Nakandala, Yuhao Zhang, Side Li, Advitya Gemawat, and Kabir Nagrecha. CIDR 2021. The CNN Hip Accelerometer Posture (CHAP) Method for Classifying Sitting Patterns from Hip Accelerometers: A Validation Study. Mikael Anne Greenwood-Hickman, Supun Nakandala, Marta M. Jankowska, Fatima Tuz-Zahra, John Bellettiere, Jordan Carlson, Paul R. Hibbing, Jingjing Zou, Andrea Z. LaCroix, Arun Kumar, and Loki Natarajan. Medicine and Science in Sports and Exercise Journal, 2021. Application of Convolutional Neural Network Algorithms for Advancing Sedentary and Activity Bout Classification. Supun Nakandala, Marta Jankowska, Fatima Tuz-Zahra, John Bellettiere, Jordan Carlson, Andrea LaCroix, Sheri Hartman, Dori Rosenberg, Jingjing Zou, Arun Kumar, and Loki Natarajan. Journal for the Measurement of Physical Behaviour, 2021. Cerebro: A Data System for Optimized Deep Learning Model Selection. Supun Nakandala, Yuhao Zhang, and Arun Kumar. VLDB 2020. Panorama: A Data System for Unbounded Vocabulary Querying over Video. Yuhao Zhang and Arun Kumar. VLDB 2020. Vista: Optimized System for Declarative Feature Transfer from Deep CNNs at Scale. Supun Nakandala and Arun Kumar. ACM SIGMOD 2020.

Advanced Database and Intelligence Lab (ADIL) is a lab under Amarnath Gupta in the San Diego Supercomputer Center (SDSC).

Admiralty federation tool - federation between kubernetes clusters

NVIDIA Aerial CUDA-Accelerated RAN and Aerial Omniverse Digital Twin

We are creating an AI surrogate models for accelerating the transport simulation problems.

An exploratory research project which aims to develop fast molecular dynamics models for drug discovery with AI. We plan to develop novel AI methods to accelerate drug design and synthesize.

Publications: LIMO: Latent Inceptionism for Targeted Molecule Generation Peter Eckmann, Kunyang Sun, Bo Zhao, Mudong Feng, Michael Gilson, Rose Yu International Conference on Machine Learning (ICML), 2022

The Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship, a program of Schmidt Sciences at UC San Diego led by Faculty Director Tara Javidi and Education Associate Director Ilkay Altıntaş, leverages UC San Diego’s place at the forefront of Artificial Intelligence research to train the next generation of scientific leaders pioneering the use of AI in STEM. Established with the generous support of Schmidt Sciences, a philanthropic initiative of Eric and Wendy Schmidt, this prestigious program annually supports 10-20 Postdoctoral Fellows with a two-year fellowship as they learn and apply AI techniques to their research in the engineering, mathematical, and natural sciences. The program provides a unique, cross-campus ecosystem of training and scientific discovery consisting of three interconnected core components: Community formation of a diverse cohort of researchers and scholars from across the natural and engineering sciences with an interest in the use of AI methods to accelerate scientific discoveries; Training modules in the form of courses, seminars, and panels pulling from established pedagogical areas of excellence designed to provide postdoctoral fellows with core AI technical competencies; and Co-mentoring to bridge the gap between STEM and AI, including at least one STEM faculty as the primary mentor and nominator, and a second AI/method faculty in the role of co-mentor. Schmidt AI in Science Postdocs also receive comprehensive foundational training in scientific leadership, research ethics, professional development, and other critical skills to complement the Program’s AI curriculum. In addition, they are encouraged to engage with the DEI initiative or program of their choice at UC San Diego.

This project aims to explore the integration of deep-learning-based facial expression recognition technology to enhance the capabilities of AI tutoring systems. Our team has developed a robust facial expression recognition model capable of identifying emotional and cognitive states in real-time. By integrating it with an AI tutoring system, we will be able to track students' emotional and cognitive responses during learning, such as engagement, confusion, and frustration. This approach enables the AI tutor to dynamically adapt its teaching strategies, pacing, and interaction style based on the student's emotional and cognitive states. For example, if the system detects confusion or frustration, the AI tutor can slow down the pace and proactively ask the student whether they need further clarifications or assistance. Through rigorous testing, we aim to demonstrate that integrating facial expression recognition systems into AI tutors can lead to more personalized, effective, and empathetic learning experiences, ultimately improving student outcomes and satisfaction

This is the AIEA lab (led by Leilani Gilpin) namespace for the auditors subgroup.

This is the AIEA lab (led by Leilani Gilpin) namespace for the interns subgroup.

This is the namespace for the AIEA lab (led by Leilani Gilpin) and the slugbotics club collaboration.

Publications: https://arxiv.org/abs/2409.10532

AI For Mankind AI For Mankind is a 501(c)(3) nonprofit organization with the mission of mobilizing the tech community to work on world challenging problems using AI and Data. We organize tech talks, workshops, and hackathons. We want to build a grassroot community of volunteers creating solutions using AI and Data to bring positive impacts to society at large.

Namespace for projects in Prof Alon Orlitsky's lab. The current focus of these projects includes large language models.

Application-Layer Traffic Optimization (ALTO) IETF workgroup

2021-2022 Amaro Lab Spike Project.This project aims to bring the machine learning and data science principles gained in the Master’s of Advanced Studies in Data Science and Engineering program to the Amaro Lab’s SARS-CoV-2 spike simulation data. The goal is to develop a machine learning model to predict whether a spike protein is in the open or closed state, and then to visualize the most important features in the model to gain insight into which protein substructures are relevant to infection of human cells. Knowledge gleaned from this process could be used not only to gain a better understanding of the dynamics of the spike protein itself but also to identify potential targets for drug development in the treatment of the SARS-CoV-2 virus.

AI and games research, primarily exploratory usage for now. (updated 1/13/23)

Publications: None yet. (updated 1/13/23)

International Research and Education Network Connections Core program: Americas-Africa Lightpaths Express and Protect (AmLight-ExP) project

Publications: https://ciara.fiu.edu/publications.html

University of California Santa Cruz Applied Machine Learning Lab

Modelling of gravitational lensing data. This includes both light curves (from LSST and other surveys) of lensed quasars and supernovae, but also modelling of imaging data of galaxy-galaxy lenses.

Machine learning to model time series of quasar variability. This involves training neural networks to fit multivariate time series with noisy and irregularly sampled data as well as perform parameter inference.

I will use the NRP for bioinformatics analyses for the NSF/FAPESP (jointly funded) “Dry Diagonal Dimensions Project” (https://www.brdd.ib.unicamp.br/). Specifically, I will use GPU nodes to run the program Cactus (https://github.com/ComparativeGenomicsToolkit/cactus) to align whole genomes of various species of lizards and snakes. Multispecies whole genome alignment is computationally demanding although other researchers have had success using the GPU-version of Cactus for this task. I do not have access to any other (non-NRP) GPUs for this analysis. Aligned genomes are necessary for functional and structural genome annotation, phylogenetic inference, and genome scans to identify genomic regions under selection and adaptive for arid environments. I only anticipate needing NRP resources for the genome alignment step.

Using the Nautilus cluster to assemble and align whole genomes

Namespace for AMNH Jupyterhub hosted on Nautilus. Used for testing by researchers

Using for Anshul for debugging the PyTorch and running code

Managed by Anthony Furman as a part of HARE Labs at UCSC.

scratch pad for development of various lab projects.

Large scale virtual screening of compounds. Indentification of first-in-lead compounds for different diseases.

ArcGIS Enterprise for Kubernetes deployment and config

A namespace for the ct-image segmentation project for surgical scene understanding.

ARchitectures for DAta group at UC Irvine. Specialized Architectures for Graph Neural Networks, Genome Sequence Alignment, Graph Mining, Precision Agriculture, and more.

argocd namespace for testing of ci/cd on gitlab services in the nautilus cluster

namespace argocd where all ArgoCD resources will be installed

Namespace for project in Prof Arya Mazumdar's lab. The main focus of projects include statistical estimation, theoretical machine learning and federated learning.

Assel's new namespace, can be used for training and testing neural networks.

Testing in-network caches for ATLAS conditions and CVMFS.

This project focus on implementation of charged particle tracking pipeline as a Triton Inference Server. Clients implemented in ACTS will send track-finding requests to the Triton server and the server will return track candidates to the client after processing. The pipeline contains several track reconstruction algorithms. Because of the heterogeneity and dependency chain of the pipeline, we will explore different server settings to maximize the throughput of the pipeline, and we will study the scalability of the inference server and time reduction of the client.

Authentik is a OAUTH provider that will proxy CiLogon for Nautilus

Autoslug is a UCSC computational club focusing on ML and Computer Vision techniques.

namespace for developing workflow for training models for cms anomaly detection at the L1 trigger (AXOL1TL)

Publications: https://cds.cern.ch/record/2876546?ln=en

A place for CSUB Staff to explore NRP and Nautilus.

Working on Variant Calling for Long sequence gene data

Publications: None.

Using neural networks to detect binary black hole mergers from time domain gravitational wave strain

Underwater ocean acoustic modelling using ray tracing on Earth sized scales.

We study biologically plausible neural networks and learning rules

Publications: To be updated

testing ssl binderhub extentions. We have binderhub instances running at SSL and want to deploy notebook servers in Nautilus.

Namespace dedicated to building software for the biocore github organization. Biocore stands for Collaboratively developed bioinformatics software.

Exploring the potential for AI to improve biodiversity research

bowtie2-genomix is part of the CPU validation of the genimix aligner

The Braingeneers are developing the infrastructure to grow cortical organoids at scale and interface with them in order to record and stimulate neurons. This will enable the application of modern AI approaches to uncover how genetic changes enhanced human brain architecture and computing capacity during primate evolution as well as to better understand how neurons function towards back porting this into in-silico machine learning models.

Publications: https://www.biorxiv.org/content/10.1101/2021.07.29.453595v2 https://cenic.org/blog/prp-boosts-inter-campus-collaboration-on-brain-research https://www.nature.com/articles/s41593-024-01715-2 https://www.biorxiv.org/content/10.1101/2024.03.15.585237v1 DOI: 10.1371/journal.pone.0312438 DOI: 10.1101/2024.11.14.623530 DOI: 10.1101/2024.11.13.623525 DOI: 10.1101/2024.03.15.585237 DOI: 10.1016/j.celrep.2023.112318 DOI: 10.1016/j.heliyon.2022.e11596 DOI: 10.1016/j.iot.2022.100618 DOI: 10.1088/1741-2552/ac310a

Experiments for braTS Segmentation Dataset for MRI Dataset. Usually requires higher-end GPUs

Neural Network Quantization using Pytorch Brevitas library. Gradual removal of skip connections through careful knowledge distillation.

Publications: https://arxiv.org/abs/2102.01351

Sensing and plant phenotyping group at University of Nebraska-Lincoln

Computer vision and robotics research for the Berkeley Vision Lab

A project focused on sustainability of computer systems design, both in terms of operational and embedded carbon reduction (https://c3lab.net)

A project focused on sustainability of computer systems design, both in terms of operational and embedded carbon reduction (https://c3lab.net)

Limited deployment of Ark nodes within the current set of unique AS represented within the cluster

Research related to biology and ecology lab at California Polytechnic State University

Publications: https://orcid.org/0000-0002-2031-2752

Namespace for use by the Computational Chemistry Group at Cal Poly San Luis Obispo

The 3D Digital Herbarium is an innovative educational platform created by the Cal Poly Humboldt Library dedicated to bringing the intricate world of botany to life through state-of-the-art 3D modeling. At the heart of our mission is the desire to transform how students learn about flora, transcending traditional boundaries by offering an immersive, interactive experience. Our 3D Digital Herbarium is a unique resource, meticulously designed for botany students and enthusiasts alike. It features a diverse collection of flora, each represented in stunning three-dimensional detail. These models offer an unparalleled opportunity to study and appreciate the intricate structures and characteristics of various plant species, providing a level of detail that far surpasses what's available in textbooks or two-dimensional images.

Publications: https://3dherbarium.org/about

To train (fine-tune or query) an open language model with Humboldt (and areas around) data, and institutional datasets to utilize as for chat and other applications. Alignment Data not limited to Digital Commons & Special Collections data sources.

James is using Jupyter for instruction. He is teaching several classes using Python in Jupyter notebooks.

Publications: http://gsp.humboldt.edu/JimsProfessional/Publications.html

Bethany is using JupyterHub for instruction. She is teaching several classes using Python in Jupyter notebooks.

LLM research, butterfly effects of early prompts, cascading design and alignment.

Microglia are a special type of immune cell found only in the central nervous system. These multifaceted cells fight infections, repair damage, remove debris, and are central to maintaining brain health. We develop an automated system to quantify microglia in the brain using computer vision.

Publications: https://sites.google.com/humboldt.edu/kamilalarripa/research?authuser=0

Reinforcement Learning experiment to improve the training efficiency of an AI model.

Publications: https://www.humboldt.edu/mathematics/rosanna-overholser

Test namespace for Cal Poly Humboldt and learning about the Nautilus Namespaces

This project seeks to design new DL algorithms for phylogenetics. We have already created a method called DEPP to update an existing tree using embeddings in Euclidean space. We are now working on methods for creating embeddings in hyperbolic spaces and divide-and-conquer methods for inferring species trees using deep learning.

Publications: Jiang, Yueyu, Metin Balaban, Qiyun Zhu, and Siavash Mirarab. “DEPP: Deep Learning Enables Extending Species Trees Using Single Genes.” Edited by Claudia Solis-Lemus. Systematic Biology, April 29, 2022, 2021.01.22.427808. https://doi.org/10.1093/sysbio/syac031. Jiang, Yueyu, Puoya Tabaghi, and Siavash Mirarab. “Phylogenetic Placement Problem: A Hyperbolic Embedding Approach.” In Comparative Genomics, edited by Lingling Jin and Dannie Durand, 68–85. Cham: Springer International Publishing, 2022. https://doi.org/10.1007/978-3-031-06220-9_5.

Server-based simulation cluster for Carla

The Collaboration for Astronomy Signal Processing and Electronics Research

This namespace is used for testing casper toolflow.

We will use this namespace for training multi-agent perception models for efficient robust collaborative perception

Publications: https://scholar.google.com/citations?user=Pni_ugMAAAAJ&hl=en&oi=ao

Exploring Cosmic Ray App (CREDO) image data with Kubernetes Federated AI Technology Enabler (KubeFATE)

Publications: None

Container Data Importer for KubeVirt

The Discovery Program at the College of Computing, Data Science, and Society incubates and accelerates high-impact research in academic, government, non-profit, and industry projects worldwide while providing UC Berkeley students with real-world research experiences and mentorship opportunities. Students gain access to advanced computing resources, data science tools, and collaborative platforms to facilitate their work. Ultimately, the Discovery Program serves as a bridge between academia and real-world applications, fostering an ecosystem where students, mentors, and external partners collaborate to produce transformative research with global impact.

The Discovery Program bridges academia and real-world applications, fostering an ecosystem where students, mentors, and external partners collaborate to produce transformative research with global impact.

CENIC SDN

JetStack cert manager deployment - provides automated certificate generation

Neural Networks applied to Smart Connected Workers. This is an edge intelligent platform to integrate internet-of-things technologies with computing hardware, software, computational workflows for machine learning, and data ingestion, enabling SMMs to transition into smart manufacturing paradigms by leveraging the intelligence of their people. The platform leverages consumer-grade electronics and sensors (affordable and portable), customized software with open-source software packages (accessible), and existing communication network infrastructures (scalable). The project utilizes Nautilus Kubernetes Cluster. The software systems are implemented via Kubernetes orchestration of Docker containerization to ensure scalability and programmability.

Publications: SCW Journal Publication Part 1: Shijie Bian, Chen Li, Yongwei Fu, Yutian Ren, Tongzi Wu, Guann-Pyng Li, and Bingbing Li*, “Machine Learning-based Real-time Monitoring System for Smart Connected Worker to Improve Energy Efficiency”, Journal of Manufacturing Systems, 2021, Vol. 61: 66-76. https://doi.org/10.1016/j.jmsy.2021.08.009 SCW Journal Publication Part 2: Yoon Kim, Richard Donovan, Yutian Ren, Shijie Bian, Tongzi Wu, Shweta Purawat, Anthony Manzo, Ilkay Altintas, Bingbing Li, and Guann-Pyng Li*, “Smart Connected Worker Edge Platform for Smart Manufacturing: Part 1: Architecture and Platform Design”, Journal of Advanced Manufacturing and Processing, 2022, Vol. 4 (4): e10129. https://doi.org/10.1002/amp2.10129 SCW Journal Publication Part 3: Richard Donovan, Yoon Kim, Anthony J Manzo, Yutian Ren, Shijie Bian, Tongzi Wu, Shweta Purawat, Henry Helvajian, Marilee Wheaton, Bingbing Li, and Guann-Pyng Li*, “Smart Connected Worker Edge Platform for Smart Manufacturing: Part 2: Implementation and On-site Deployment Case Study”, Journal of Advanced Manufacturing and Processing, 2022, Vol. 4 (4): e10130. https://doi.org/10.1002/amp2.10130 SCW Journal Publication Part 4: Chen Li, Shijie Bian, Tongzi Wu, Richard P. Donovan, and Bingbing Li*, “Affordable Artificial Intelligence-Assisted Machine Supervision System for Small and Medium-Sized Manufacturers”, Sensors (JCR2024 Impact Factor: 3.4), 2022, Vol. 22 (16): 6246. https://doi.org/10.3390/s22166246 SCW Journal Publication Part 5: Bingbing Li, Tongzi Wu, Shijie Bian, John W. Sutherland*, “Predictive Model for Real-time Energy Disaggregation Using Long Short-term Memory”. CIRP Annals - Manufacturing Technology, 2023, Vol. 72 (1): 25-28. https://doi.org/10.1016/j.cirp.2023.04.066 SCW Conference Publication Part 1: Shijie Bian, Tiancheng Lin, Chen Li, Yongwei Fu, Mengrui Jiang, Tongzi Wu, Xiyi Hang, and Bingbing Li*, “Real-time Object Detection for Smart Connected Worker in 3D printing”, Proceedings of the 2021 International Conference on Computational Science (ICCS 2021), Krakow, Poland, June 16-18, 2021. https://doi.org/10.1007/978-3-030-77970-2_42

Using structure from motion (SFM), we are able to reconstruct dense 3D point cloud models of coral reefs from photographic surveys. We are leveraging advances in 2D and 3D computer vision to increase the speed and accuracy of the annotation of these models for use in ecological and biological research.

Publications: https://www.frontiersin.org/journals/robotics-and-ai/articles/10.3389/frobt.2022.884317/full

modeling chess decision-making via simulations and informational principles

The Chodera lab uses computation and experiment to develop quantitative, multiscale models of the effects of small molecules on biomolecular macromolecules and cellular pathways and understand the functional and therapeutic ramifications of mutations. The group utilizes physical models, rigorous statistical mechanics, and open source software development practices with overall goals of engineering novel therapeutics and tools for chemical biology, predicting resistance or susceptibility to therapy, and understanding the physical driving forces behind the emergence of drug resistance. We develop and use advanced algorithms for molecular dynamics simulations on GPUs and distributed computing platforms, in addition to high-throughput experiments to characterize biophysical interactions between small molecules and their targets.

Publications: "None"

Intersubject variability among individuals who use opioids is evident, driven by the neuroplastic changes resulting from chronic opioid use. These changes give rise to a range of affective states, including heightened stress, increased pain sensitivity, and the emergence of opioid cravings. Monitoring these affective states is critical to screen and alert at-risk users who are susceptible to developing opioid use disorder. While mobile and wearable sensor devices have demonstrated their ability to model these states broadly, they often fall short in effectively adapting to the nuanced intersubject variability observed in these changes. In response, we propose a hierarchical deep learning approach capable of dynamically identifying optimal user group segmentation for personalized prediction of levels of stress, pain, and craving based on heart rate variability data from a wearable wristband. We trained and evaluated our methods on a dataset collected from 51 subjects and around 2,000 clean samples, each of which contains hours of heart rate variability data and EMA responses. The empirical results show that the hierarchical model structure can significantly boost the performance by leveraging information from the physiological and demographic attributes.

Used to host short workshops on research computing with kubernetes and the NRP.

Compute and storage in support of the Climate Analytics Lab, including large-scale inference over satellite imagery and climate model emulation.

Publications: - https://www.pnas.org/doi/full/10.1073/pnas.2206885119 - https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2021MS002954 - etc.

Advancing climate science through innovative applications and fundamental developments of machine learning research. For example, generating more efficient and accurate weather forecasts and climate projections with skillful uncertainty quantification.

Publications: DYffusion: A Dynamics-informed Diffusion Model for Spatiotemporal Forecasting, Salva Rühling Cachay, Bo Zhao, Hailey Joren, and Rose Yu, Advances in Neural Information Processing Systems (NeurIPS), 2023

Clockwork is the high-frequency time synchronization software alowing to achive sub-millisecond accuracy over long distance

Working towards large attention model development for vision task.

Cluster topology system namespace is the deployment for nebula studio graph database to support the traceroute tool

Namespace dedicated to CMS services, on CMS hardware. A portion of the SoCal Cache runs here.

CMS ML activities, including machine-learned particle-flow reconstruction, particle graph autoencoders, autoencoders for data compression, model pruning/quantization/compression studies, ML for analysis (Higgs to WW, and LLP jet tagging)

Publications: Accelerated Charged Particle Tracking with Graph Neural Networks on FPGAs [https://arxiv.org/abs/2012.01563], MLPF: Efficient machine-learned particle-flow reconstruction using graph neural networks [https://arxiv.org/abs/2101.08578], Graph Generative Adversarial Networks for Sparse Data Generation in High Energy Physics [https://arxiv.org/abs/2012.00173], The LHC Olympics 2020: A Community Challenge for Anomaly Detection in High Energy Physics [https://arxiv.org/abs/2101.08320], Ps and Qs: Quantization-aware pruning for efficient low latency neural network inference [https://arxiv.org/abs/2102.11289], Explaining machine-learned particle-flow reconstruction [https://arxiv.org/abs/2111.12840], Particle Cloud Generation with Message Passing Generative Adversarial Networks [https://arxiv.org/abs/2106.11535], Particle Graph Autoencoders and Differentiable, Learned Energy Mover's Distance [https://arxiv.org/abs/2111.12849], Machine Learning for Particle Flow Reconstruction at CMS [https://arxiv.org/abs/2203.00330], hls4ml: An Open-Source Codesign Workflow to Empower Scientific Low-Power Machine Learning Devices [https://arxiv.org/abs/2103.05579], A reconfigurable neural network ASIC for detector front-end data compression at the HL-LHC [https://arxiv.org/abs/2105.01683], A FAIR and AI-ready Higgs boson decay dataset [https://arxiv.org/abs/2108.02214]

For Higgs bosons -> Vector bosons graph neural network classifier development.

Publications: Search for nonresonant pair production of highly energetic Higgs bosons decaying to bottom quarks [https://arxiv.org/abs/2205.06667], More CMS publications in preparation

CMS users at UCSD.

Coder shifts software development from local machines to on-prem and public cloud infrastructure. Onboard new developers in minutes, build code on powerful servers—all while keeping source code and data secure behind your firewall.

Coder for UCSD and SDSC dev and networking. Similar to the coder namespace but not open.

Publications: None

Setup by Ashton to use for users needing space for developing images, particularly for docker in docker

Nautilus access for bird repopulation research and networking lab research, XAI testing, ACM Research Lab (PINN and medical imagery explainable classification)

Publications: "None"

For use by the students of Cognitive Robotics laboratory under Prof. Henrik I. Christensen.

Working on eye-tracking for image and video compression and its impact

As more and more AI agents are used in practice, it is time to think about how to make these agents fully autonomous so that they can (1) learn by themselves continually in a self-motivated and self-initiated manner in the open world environment rather than being retrained offline periodically on the initiation of human engineers and (2) accommodate or adapt to unexpected or novel circumstances. As the real-world is an open environment that is full of unknowns or novelties, the ability to detect novelties (out-of-distribution cases), characterize them, accommodate/adapt to them, and gather ground-truth training data and continually or incrementally learn the unknowns/novelties is becoming critical in making the AI agent more and more knowledgeable, powerful, and self-sustainable over time. Our lab works on these challenging problems.

Experiments on algorithms related to cooperative perception.

Coturn installation - the server passing through WebRTC audio and video traffic

The Cyber-Physical Networking Lab is an international research hub dedicated to innovations in wireless networks of systems that are aware of, can timely adapt to, and change their environment. We focus on challenging problems at the intersection of wireless communications, AI, networking, and the physical world, supported by grants from the National Science Foundation, the Department of Defense, the Department of Energy, and the Department of Transportation. CPN Lab alumni are employed at Purdue University, Rice University, Apple, Amazon, Qualcomm, NVidia, and Microsoft.

Publications: https://cpn.unl.edu/publications/

Namespace for Dr. Huseyin Ucar at Cal Poly Pomona.

CRXEL focuses on grassroots creativity and learning while promoting research that combines artificial intelligence with interactive and participatory media to allow people to be more creative, informed and make better decisions in educational as well as entertainment environments

Research on the propagation of birds. Joint research between Computer Science and Ecology.

Research and teaching for Culver-Stockton College in Canton, MO

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottg

Research and teaching for Culver-Stockton College in Canton, MO

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottg

This is a namespace for analyzing the energy footprint of executing compute jobs. It is a part of the CSE 297 Fall 2024 Class at UCSC/CSE.

Publications: https://asouza.io/publications

The Technology Infrastructure for Data Exploration (TIDE) is designed to support cutting-edge research in machine learning and AI. The computing architecture is based on powerful graphics, high-performance processors, and ample storage to support research discovery. This JupyterHub instance acts as the front door for CSU researchers to the TIDE cluster.

This is the CSU TIDE Jupyterhub development environment

Focus on providing a proof of concept for the CSUF account course.

Testing Hadoop with clustering for CPSC 531 - Advanced Databases

General workspace environment for testing projects

Focus on testing new images and new concepts for windows vm.

California State University Fullerton faculty sandbox environment.

Focus on providing a proof of concept CSUF teaching and learning space.

Focus on providing a proof of concept CSUF teaching and learning space.

Will be used for JupyterLab research for faculty and staff.

Focus on providing a proof of concept CSUF teaching and learning space.

Development JupyterHub environment for research and instructional use.

Production JupyterHub environment for research and instructional use.

This cluster was created for testing purposes to see how to support our research staff with NRP.

Test at Cal State Monterey Bay. I am trying to run the basic kubernetes tutorial pod

California State University Northridge’s (CSUN) Autonomy Research Center for STEAHM (ARCS) is a NASA-sponsored, chartered, Center of Excellence. Our mission is to combine transdisciplinary, university-wide knowledge and talent from faculty, students, and NASA scientists to conduct convergence research and collaboration using increasingly autonomous systems (IA).

Publications: Journal Publication Part 1: Shijie Bian, Daniele Grandi, Tianyang Liu, Pradeep K. Jayaraman, Karl Willis, Elliot Sadler, Bodia Borijin, Thomas Lu, Richard Otis, Nhut Ho, and Bingbing Li*, “HG-CAD: Hierarchical Graph Learning for Material Prediction and Recommendation in CAD”, Journal of Computing and Information Science in Engineering, 2024, Vol. 24 (1): 011007. https://doi.org/10.1115/1.4063226 Conference Publication Part 1: Shijie Bian, Daniele Grandi, Kaveh Hassani, Elliot Sadler, Bodia Borijin, Axel Fernandes, Andrew Wang, Thomas Lu, Richard Otis, Nhut Ho, Bingbing Li*, “Material Prediction for Design Automation Using Graph Representation Learning”, Proceedings of the ASME 2022 International Design Engineering Technical Conferences &Computers and Information in Engineering Conference (IDETC/CIE 2022), St. Louis, Missouri, U.S.A., August 14-17, 2022. https://doi.org/10.1115/DETC2022-88049 Conference Publication Part 2: Anthony Morales-Badajoz, Neville Elieh, April Diederich, Elliot Sadler, Jasmine Glover, Manoj Nizampatnam, Troy Israel, Andrew Wang, Larry Zhang, Annette Besnilian, Andreas George, Julie Miller, Xunfei Jiang, Bingbing Li*, “Astro Cultivators: Autonomous Growth System for Space Farming based on Machine Vision and Multi-Sensor Fusion”, Proceedings of ACM Cyber-Physical Systems and Internet of Things Week 2023 (CPS-IoT Week Workshops '23), San Antonio, Texas USA, May 9-12, 2023. https://doi.org/10.1145/3576914.3588338

We are mainly focused on developing deep learning models for the following image processing tasks. (1) Image denoising. Convoluion neural networks , vision transformers, and the hybrid models are investigated for image denoising. (2) Image reconstruction. Convoluion neural networks , vision transformers, and the hybrid models are investigated for CT and MRI image reconstruction.

Publications: “Real-time Object Detection for Smart Connected Worker in 3D printing” by Shijie Bian, Tiancheng Lin, Chen Li, Yongwei Fu, Mengrui Jiang, Tongzi Wu, Xiyi Hang, Bingbing Li*. Proceedings of the 2021 International Conference on Computational Science (ICCS 2021), Krakow, Poland, June 16-18, 2021.

We are mainly focused on developing deep learning models for the following image processing tasks. (1) Image denoising. Convoluion neural networks , vision transformers, and the hybrid models are investigated for image denoising. (2) Image reconstruction. Convoluion neural networks , vision transformers, and the hybrid models are investigated for CT and MRI image reconstruction.

Publications: Real-time Object Detection for Smart Connected Worker in 3D printing” by Shijie Bian, Tiancheng Lin, Chen Li, Yongwei Fu, Mengrui Jiang, Tongzi Wu, Xiyi Hang, Bingbing Li*. Proceedings of the 2021 International Conference on Computational Science (ICCS 2021), Krakow, Poland, June 16-18, 2021.

This namespace is dedicated to AI research and training at CSUSB

JupyterHub for Dr. Jeremy Aikin at Cal State San Bernardino

xREAL is an interdisciplinary technology innovation hub that brings together faculty, students, staff, and industry partners and uses a variety of leading-edge technologies to design and develop immersive learning experiences that advance the scholarship of teaching and learning and have demonstrable pedagogical benefits. Our mission is to transform teaching and learning with leading-edge technologies by researching and designing surprising human-machine interactions that encourage the joy of discovery, provide educational insights, and contribute to the public good. Our mission is to transform teaching and learning with leading-edge technologies by researching and designing surprising human-machine interactions that encourage the joy of discovery, provide educational insights, and contribute to the public good.

Publications: https://www.csusb.edu/academic-technology-innovation/xreal-lab

The Vienna Ab initio Simulation Package (VASP) is a computer program for atomic scale materials modeling, e.g. electronic structure calculations and quantum-mechanical molecular dynamics, from first principles. https://www.vasp.at/index.php/about-vasp/59-about-vasp

Publications: "None"

This namespace is for Dr. Kimberley Cousins at Cal State San Bernardino

This namespace is for TIDE grad assistant to develop and deploy software

The namespace will be used for Dr. Hamouda's F2022 course, IST 6110 & 6620 Workshop Research at Cal State San Bernardino.

Publications: "None"

This namespace is for JupyterHub dedicated to faculty research at California State University, San Bernardino

Publications: "None"

This namespace is for JupyterHub and stack developments and collaboration with other California State Universities

dedicated to AI supported Entrepreneurship research

This name space is dedicated to Prof. Jennifer Jin and her students in ML/AI research

This namespace is for JupyterHub dedicated to the members of California State University, San Bernardino for classroom and research

Publications: "None"

Professor Khan's research lab in AI/ML and computer vision

JupyterHub for Spring 2025 classes for Youngsu Kim

This project utilizes electronic health records (EHR) mostly available from public datasets, such as the Agency for Healthcare Research and Quality's (AHRQ) Healthcare Cost and Utilization Project (HCUP) or California Office of Statewide Health Planning and Development (OSHPD) to uncover predictive patterns in patient or hospital level health-related outcomes.

Publications: https://www.csusb.edu/academic-technologies-innovation/xreal-lab-and-high-performance-computing/high-performance-computing/projects

This namespace is dedicated to nautilus training at CSUSB

This namespace is for JupyterHub that will be used for CSUSB BIOL-5050, Dr. Bree Putman, in Fall 2023

This namespace is dedicated to research Information & Decision Sciences Department and collaborations with other institutions

for Professor Qiuxiao Chen's research in computer vision

The namespace is for Summer 2023 research at CSUSB

The Vienna Ab initio Simulation Package (VASP) is a computer program for atomic scale materials modelling, e.g. electronic structure calculations and quantum-mechanical molecular dynamics, from first principles. https://www.vasp.at/index.php/about-vasp/59-about-vasp

Publications: Functional materials are extended solids that respond to external stimuli such as electric or magnetic fields, light, or heat, by changing structure/polarity/magnetic anisotropy, and which may be tuned at the molecular level. The Center for Advanced Functional Materials at CSUSB, funded in part by an NSF-CREST grant (NSF 1914777) seeks to develop, discover, and apply new functional materials in crystalline, thin film, and polymeric forms. Dr. Cousins, and new faculty member Dr. Joyce Pham, contribute to this endeavor computationally using density functional theory, and molecular dynamics will be used to describe and study materials at the molecular level. A fundamental question that a computational investigation may help answer is “what gives rise to the observed structures,” which thus provides an avenue to understand and better design a material with desired properties. The insight gained from computation helps us optimize existing materials and predict new materials for study by our experimental colleagues. 1. Kimberley Cousins & Sarah Rodriguez* “Materials genome approach to functional materials discovery using the CSD” at the One Million Crystal Structures Symposium (oral presentation), American Chemical Society Fall 2019 National Meeting, San Diego, CA, August 26, 2019. Re-reported in a white paper by Wendy Warr, November, 2019. 2. Timothy Usher, Kimberley Cousins, Douglas Smith, Renwu Zhang, Eva Zurek, Stephen Ducharme, Sara Callori, Daniel Miller**, Paulo Costa**. “Materials Genome Approach to Organic Ferroelectrics and Piezoelectrics.” Int. J. Nanotechnol., Vol. 15, Nos. 8/9/10, 2018. 3. Bindi, L.; Pham, J.; Steinhardt, P.J. “Previously Unknown Quasicrystal Periodic Approximant Found in Space.” Scientific Reports 2018 (8), 16271, 1–7. 4. Pham, J.; Miller, G.J. “AAuAl (A = Ca, Sc, and Ti): Peierls Distortion, Atomic Coloring, and Structural Competition.” Inorganic Chemistry 2018 57 (7), 4039–4049. (Open access link via Ames Lab accepted manuscript digital repository) 5. Miller, G.J.; Pham. J.; Smetana, V.; Xie, W. “Unraveling Complex Intermetallics Using a Cluster Perspective.” Springer-Nature: Chapter Contribution to Commemorate Ken Wade’s 50-Year Anniversary Publication on the ‘Wade’s Rules’. Available 2021.

xREAL is an interdisciplinary technology innovation hub that brings together faculty, students, staff, and industry partners and uses a variety of leading-edge technologies to design and develop immersive learning experiences that advance the scholarship of teaching and learning and have demonstrable pedagogical benefits.

Publications: https://www.csusb.edu/academic-technologies-innovation/xreal-lab-and-innovation

For f2024 math2265 at California State University San Bernardino

Advanced AI Solutions for Healthcare and Decision Support for Prof. Yan Zhang

Fake news detection and data analysis at CSU Stanislaus.

A namespace for research for ML Computer Vision lab Namespace at CSU Stanislaus.

A Helm chart for the CVMFS-CSI driver, allowing the mounting of CVMFS repositories in Kubernetes environments. This chart will deploy the CSI driver as a DaemonSet, thus automatically scaling the driver on each cluster node.

this is for CXL deivce application basad profiling, test, and managment.

This projects end goal is to create an immersive Cyber-Archaeologists' Warehouse inside UE5 for analysis and study of digitized assets from field excavations and survey. We see this as the next step to digitization, analysis and collaboration that allows us to begin building out the metaverse of the archaeological past.

Dask operator provides the management for dask clusters in kubernetes

end to end testing of LIGO's data analysis workflow.

Publications: None for specific project

This project aims to develop a machine learning model capable of detecting and categorizing defects in printed circuit boards (PCBs). The model will work with high-resolution images of the MAO and SCI4 boards to identify if a given fault is a "killer" defect or a benign one. Two primary approaches are in consideration: a supervised model trained on a limited set of both defect types and an image segmentation model to detect abnormalities relative to critical components.

consistent dataset distillation (intern project) studies whether the current DD methods offers what they promise

Design and testing of machine learning approaches to translate classical and academic Chinese into academic English.

Combining Simulated and Real Data for Near-Term Forecasting of Nonstationary Dynamic Processes, with applications to Traffic and COVID-19 forecasting. Developing hybrid physics-guided deep learning tools to forecast non-stationary, non-linear dynamics

Publications: Bridging Physics-based and Data-driven modeling for Learning Dynamical Systems Rui Wang, Danielle Maddix, Christos Faloutsos, Yuyang Wang, Rose Yu Annual Conference on Learning for Dynamics and Control (L4DC), 2021 Traffic Forecasting using Vehicle-to-Vehicle Communication Steven Wong, Lejun Jiang, Robin Walters, Tamás G. Molnár, Gábor Orosz, Rose Yu Annual Conference on Learning for Dynamics and Control (L4DC), 2021

This namespace is created for research work under Rose group (deep-forecast) for point-process related research

Publications: [1] Zhou, Z. & Yu, R., Automatic Integration for Fast and Interpretable Neural Point Processes., Learning for Dynamics and Control (L4DC), 2023 [2] Zhou, Z., Yang, X., He, X., Rossi, R., Zhao, H., & Yu, R., Neural Point Process for Learning Spatiotemporal Event Dynamics., Learning for Dynamics and Control (L4DC), 2022

2023 DSE260 Capstone Project: Fire Simulation Data Analysis

The Feltus lab is running deep learning oncogenomics workflows on the Pacific Research Platform Kubernetes (K8s) cluster. The PRP K8s is allowing us to scale up our analyses by moving large genomics datasets between FIONA nodes and then screening tens of thousands of genes on GPUs for tumor biomarker discovery.

Training deep learning models for analysis of large-scale structure in the spatial distribution of galaxies

default kubernetes namespace, not used

ic design tools (Cadence / Synopsys), hardware simulations

Publications: Arpan Dutta, et al.,”HDnn PIM: Efficient in Memory Design of Hyperdimensional Computing with Feature Extraction,” GLVLSI’22. Minxuan Zhou, et al., “TransPIM: A Memory-based Acceleration via Software-Hardware Co-Design for Transformers," HPCA'22.

CRISPS cell centric image similarity projection tool for visualization namespace for collaborations with ANS

This research focuses on integrating controllable diffusion models into traffic simulation. The project aims to develop a model capable of simulating diverse traffic scenarios by incorporating real-world data like vehicle density and road conditions. A significant aspect is the dynamic control of model parameters, enabling the simulation of various traffic situations, such as peak-hour congestion and emergency scenarios. The goal is to improve the capabilities of traffic simulations for applications in autonomous vehicle development.

Digester adds the image digest to all deployed pods. More info at https://github.com/XenitAB/spegel/blob/main/docs/FAQ.md

Smoke Detection (formerly NVIDIA DIGITS) - Wildfire smoke detection from images using deep learning

Publications: A. Dewangan, Y. Pande, H.-W. Braun, F. Vernon, I. Perez, I. Altintas, G. Cottrell, and M. H. Nguyen. "FIgLib & SmokeyNet: Dataset and Deep Learning Model for Real-Time Wildland Fire Smoke Detection." Remote Sensing 14, no. 4 (2022): 1007. https://www.mdpi.com/2072- 4292/14/4/1007 A. Dewangan, Y. Pande, H.-W. Braun, F. Vernon, I. Perez, I. Altintas, G. Cottrell, and M. H. Nguyen. “FIgLib & SmokeyNet: Dataset and Deep Learning Model for Real-Time Wildland Fire Smoke Detection,” in NeurIPS Workshop on Tackling Climate Change with Machine Learning, 2021. https://www.climatechange.ai/papers/neurips2021/26

Test namespace for administrative stuff, update test, update test 2

The overarching goal of our research is to effectively and efficiently discover knowledge from large amounts of digital data. In particular, we are interested in extracting information for knowledge graph construction; modeling hierarchical structures for natural language understanding; learning robust and accurate models in low resource settings.

This is a namespace dedicated to health data donated by individuals to the Smarr Lab at UCSD.

drawio

Publications: NA

Kubernettes testing and deployment for DISE. We will use this as a Dev platform

Namespace for Rankin lab used for NN model training and studies as well as tests of FPGA acceleration

Publications: https://arxiv.org/abs/2010.08556

Namespace for Dustin Wang for test system, like Transformers

Analyzing longitudinal blood-glucose levels from individuals with type 2 diabetes.

Used for the engineers for exploration FishSense project

Publications: https://ccrutchf.github.io

We integrate genomics and deep learning for ecDNA-driven cancer diagnostics.

Publications: - Chowdhry, S. et al. NAD metabolic dependency in cancer is shaped by gene amplification and enhancer remodeling. Nature (2019) - Rajkumar, U. et al. ecSeg: Semantic Segmentation of Metaphase Images containing Extrachromosomal DNA. iScience. (2019) - Turner, K. et al. Circular extrachromosomal DNA drives massive oncogene expression and chromatin remodeling. Nature. (2019) - Luebeck, J. et al AmpliconReconstructor: Integrated analysis of NGS and optical mapping resolves the complex structures of focal amplifications in cancer. bioRxiv. (2020) - Kim, H. et al. Frequent extrachromosomal oncogene amplification drives aggressive tumors, Nature Genetics (2020) - Lange et al., Principles of ecDNA random inheritance drive rapid genome change and therapy resistance in human cancers, Nature Genetics (2022) - Hung et al., EcDNA hubs drive cooperative intermolecular oncogene expression, Nature (2021)

We are working on designing ML based encoder and decoder for a communication system

Signal Transmission and Recording (STAR) group project at the Center for Memory and Recording Research (CMRR). In this project, we explore several topics in machine learning: 1) The application of machine learning to device optimization and failure prediction in storage devices. 2) The design and evaluation of neural-assisted algorithms to enhance the performance of digital communication and storage systems. 3) The sensitivity of machine learning models to errors in the model parameters, and the application of error correction coding techniques to provide robustness to such errors. 4) Generative models for simulating signals and sequences produced by magnetic recording and non-volatile memory devices. 5) Using vector index data structure to accelerate clustering of biological data.

Publications: https://cmrr-star.ucsd.edu/publications/

With continued growth in the demands on the wireless spectrum for wireless communication, spectrum policies are evolving at a pace far more rapid than ever before. Central to efforts of spectrum modernization is a critical need to accurately measure spectrum activities across diverse, wide bands and across wide areas in a cost-effective and accurate manner, so that impacts of such changes can be carefully evaluated and acted upon in a data-driven manner. The focus of this project, SpecScape, is to design, implement, deploy, and make available low-cost kits that allow spectrum sensing and measurement. In particular, the team is building an end-to-end infrastructure that includes mobile sensors to measure spectrum activity, a supporting software ecosystem, a cloud-hosted infrastructure to manage collected measurements, and mechanisms by which users can access such information. The most significant broader impact of this project is that it provides a community-driven way to understand spectrum use across different spectrum bands -- across communications, astronomy, weather prediction, localization systems, etc. This information will aid researchers, industry practitioners, and governmental agencies, including policymakers. On the educational side, the team is involved in creating a hands-on wireless curriculum for undergraduates across multiple institutions (UW and UCSD), engaging undergraduates in research-related activities, and creating online courseware based on the spectrum sensing platform. The project also engages a broad audience through multiple dissemination channels of research outcomes and aims to encourage women and minority students to pursue STEM careers through opportunities in research activities.

Publications: [1] Bansal K, Rungta K, Bharadia D. RadSegNet: A Reliable Approach to Radar Camera Fusion. arXiv preprint arXiv:2208.03849. 2022 Aug 8. [2] Dureppagari HK, Dinesha U, Wu R, Ganji S, Ko WH, Shakkottai S, Bharadia D. Realtime intelligent control for NextG cellular radio access networks. InProceedings of the 20th Annual International Conference on Mobile Systems, Applications and Services 2022 Jun 27 (pp. 567-568). [3] Givehchian H, Bhaskar N, Herrera ER, Soto HR, Dameff C, Bharadia D, Schulman A. Evaluating Physical-Layer BLE Location Tracking Attacks on Mobile Devices. In2022 IEEE Symposium on Security and Privacy (SP) 2022 May 22 (pp. 1690-1704). IEEE. [4] Arun A, Ayyalasomayajula R, Hunter W, Bharadia D. P2SLAM: Bearing Based WiFi SLAM for Indoor Robots. IEEE Robotics and Automation Letters. 2022 Jan 25;7(2):3326-33. [5] Wu Y, Ayyalasomayajula R, Bianco MJ, Bharadia D, Gerstoft P. Sound source localization based on multi-task learning and image translation network. The Journal of the Acoustical Society of America. 2021 Nov 5;150(5):3374-86. [6] Zhao M, Chang T, Arun A, Ayyalasomayajula R, Zhang C, Bharadia D. ULoc: Low-power, scalable and cm-accurate UWB-tag localization and tracking for indoor applications. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies. 2021 Sep 14;5(3):1-31. [7] Wu Y, Ayyalasomayajula R, Bianco MJ, Bharadia D, Gerstoft P. Sslide: Sound source localization for indoors based on deep learning. InICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2021 Jun 6 (pp. 4680-4684). IEEE.

Machine Learning for Physical Layer Communication, Sensing, and Inverse Problems.

This is my research group's work on AI-enabled Optimization.

With continued growth in the demands on the wireless spectrum for wireless communication, spectrum policies are evolving at a pace far more rapid than ever before. Central to efforts of spectrum modernization is a critical need to accurately measure spectrum activities across diverse, wide bands and across wide areas in a cost-effective and accurate manner, so that impacts of such changes can be carefully evaluated and acted upon in a data-driven manner. The focus of this project, SpecScape, is to design, implement, deploy, and make available low-cost kits that allow spectrum sensing and measurement. In particular, the team is building an end-to-end infrastructure that includes mobile sensors to measure spectrum activity, a supporting software ecosystem, a cloud-hosted infrastructure to manage collected measurements, and mechanisms by which users can access such information.

study on 3d presentation with self-supervision and for robotics. We will learn 3D representations from videos and use it for robotic manipulation tasks.

Publications: Rishabh Jangir*, Nicklas Hansen*, Sambaran Ghosal, Mohit Jain, Xiaolong Wang. Look Closer: Bridging Egocentric and Third-Person Views with Transformers for Robotic Manipulation. Robotics and Automation Letters (RA-L), 2022. Zihang Lai, Sifei Liu, Alexei A. Efros, Xiaolong Wang. Video Autoencoder: self-supervised disentanglement of static 3D structure and motion. International Conference on Computer Vision (ICCV), 2021 (Oral Presentation).

5G communication stack and networking development. These students will use high-performance GPUs to accelerate 5G data processing and set up standard-compliant nodes for performance testing and research.

Publications: Jain, Ish Kumar, Raghav Subbaraman, Tejas Harekrishna Sadarahalli, Xiangwei Shao, Hou-Wei Lin, and Dinesh Bharadia. "mMobile: Building a mmWave Testbed to Evaluate and Address Mobility Effects." In Proceedings of the 4th ACM Workshop on Millimeter-Wave Networks and Sensing Systems, pp. 1-6. 2020. [1] Bansal K, Rungta K, Bharadia D. RadSegNet: A Reliable Approach to Radar Camera Fusion. arXiv preprint arXiv:2208.03849. 2022 Aug 8. [2] Dureppagari HK, Dinesha U, Wu R, Ganji S, Ko WH, Shakkottai S, Bharadia D. Realtime intelligent control for NextG cellular radio access networks. InProceedings of the 20th Annual International Conference on Mobile Systems, Applications and Services 2022 Jun 27 (pp. 567-568). [3] Givehchian H, Bhaskar N, Herrera ER, Soto HR, Dameff C, Bharadia D, Schulman A. Evaluating Physical-Layer BLE Location Tracking Attacks on Mobile Devices. In2022 IEEE Symposium on Security and Privacy (SP) 2022 May 22 (pp. 1690-1704). IEEE. [4] Arun A, Ayyalasomayajula R, Hunter W, Bharadia D. P2SLAM: Bearing Based WiFi SLAM for Indoor Robots. IEEE Robotics and Automation Letters. 2022 Jan 25;7(2):3326-33. [5] Wu Y, Ayyalasomayajula R, Bianco MJ, Bharadia D, Gerstoft P. Sound source localization based on multi-task learning and image translation network. The Journal of the Acoustical Society of America. 2021 Nov 5;150(5):3374-86. [6] Zhao M, Chang T, Arun A, Ayyalasomayajula R, Zhang C, Bharadia D. ULoc: Low-power, scalable and cm-accurate UWB-tag localization and tracking for indoor applications. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies. 2021 Sep 14;5(3):1-31. [7] Wu Y, Ayyalasomayajula R, Bianco MJ, Bharadia D, Gerstoft P. Sslide: Sound source localization for indoors based on deep learning. InICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2021 Jun 6 (pp. 4680-4684). IEEE.

In healthcare, due to data privacy and security issues, it is difficult to obtain a large amount of training data. Machine learning models, especially deep learning models, typically have a lot of weight parameters. Training large-sized models on small datasets can easily lead to overfitting, meaning that the models perform well on training data but generalize poorly on unseen test data. To address this problem, I and my students have been developing sample-efficient ML methods which can train highly- performant models on small-sized medical data. While current progress in AI for healthcare is encouraging, not too many clinical AI solutions are deployed in hospitals or actively utilized by physicians. A major problem is that existing clinical AI methods are less trustworthy. For example, existing approaches make clinical decisions in a black-box way, which renders the decisions difficult to understand and less transparent. Existing solutions are not robust to small perturbations or potentially adversarial attacks, which raises security and privacy concerns. As a result, physicians are reluctant to use these solutions since clinical decisions are mission- critical and must be made with high trust and reliability. To address these problems, I and my students have been developing trustworthy ML methods for healthcare, which are interpretable and robust against adversarial attacks.

Publications: Yijian Qin, Xin Wang, Ziwei Zhang, Pengtao Xie, Wenwu Zhu. Graph Neural Architecture Search Under Distribution Shifts. International Conference on Machine Learning (ICML), 2022. Pengtao Xie and Xuefeng Du. Performance-Aware Mutual Knowledge Distillation for Improving Neural Architecture Search. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022. Yuren Mao, Zekai Wang, Weiwei Liu, Xuemin Lin, and Pengtao Xie. MetaWeighting: Learning to Weight Tasks in Multi-Task Text Classification. The 60th Annual Meeting of the Association for Computational Linguistics (ACL), Findings, 2022. Youwei Liang, Chongjian Ge, Zhan Tong, Yibing Song, Jue Wang, and Pengtao Xie. EViT: Expediting Vision Transformers via Token Reorganizations. International Conference on Learning Representations (ICLR), 2022. (Spotlight Presentation) Sai Ashish Somayajula, Linfeng Song and Pengtao Xie. A Multi-Level Optimization Framework for End-to-End Text Augmentation. Transactions of the Association for Computational Linguistics (TACL), 2022. Bhanu Garg, Li Zhang, Pradyumna Sridhara, Ramtin Hosseini, Eric Xing, and Pengtao Xie. Learning from Mistakes -- A Framework for Improving Neural Architecture Search. AAAI Conference on Artificial Intelligence (AAAI), 2022. Pengtao Xie, Jun Zhu, and Eric P. Xing. Diversity-promoting Bayesian Learning of Latent Variable Models. Conditionally accepted by the Journal of Machine Learning Research (JMLR). Jiayuan Huang, Yangkai Du, Shuting Tao, Kun Xu, and Pengtao Xie. Structured Self-Supervised Pretraining for Commonsense Knowledge Graph Completion. Transactions of the Association for Computational Linguistics (TACL), 2021. Xuehai He, Zhuo Cai, Wenlan Wei, Yichen Zhang, Luntian Mou, Eric Xing and Pengtao Xie. Towards Visual Question Answering on Pathology Images. The 59th Annual Meeting of the Association for Computational Linguistics (ACL), 2021. Meng Zhou, Zechen Li, Bowen Tan, Guangtao Zeng, Wenmian Yang, Xuehai He, Zeqian Ju, Subrato Chakravorty, Shu Chen, Xingyi Yang, Yichen Zhang, Qingyang Wu, Zhou Yu, Kun Xu, Eric Xing and Pengtao Xie. On the Generation of Medical Dialogs for COVID-19. The 59th Annual Meeting of the Association for Computational Linguistics (ACL), 2021. Ramtin Hosseini, Xingyi Yang and Pengtao Xie. DSRNA: Differentiable Search of Robust Neural Architectures. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021. Meng Zhou, Zechen Li and Pengtao Xie. Self-supervised Regularization for Text Classification. Transactions of the Association for Computational Linguistics (TACL), 2021. Jiaqi Zeng and Pengtao Xie. Contrastive Self-supervised Learning for Graph Representation Learning. AAAI Conference on Artificial Intelligence (AAAI), 2021. Seojin Bang, Pengtao Xie, Heewook Lee, Wei Wu, Eric Xing. Explaining Black-box Models Using A Deep Variational Information Bottleneck Approach. AAAI Conference on Artificial Intelligence (AAAI), 2021. Luntian Mou, Chao Zhou, Pengtao Xie, Pengfei Zhao, Ramesh Jain, Wen Gao, and Baocai Yin. Isotropic Self-supervised Learning for Driver Drowsiness Detection with Attention-based Multimodal Fusion. IEEE Transactions on Multimedia (TMM), 2021. Jeanne Vu, Ghiam Yamin, Zabrina Reyes, Alex Shin, Alexander Young, Irene Litvan, Pengtao Xie, Sebastian Obrzut. Assessment of Motor Dysfunction with Virtual Reality in Patients Undergoing [123I]FP-CIT SPECT/CT Brain Imaging. Tomography, 2021. G. Zeng, W. Yang, Z. Ju, Y. Yang, S. Wang, R. Zhang, M. Zhou, J. Zeng, X. Dong, R. Zhang, H. Fang, P. Zhu, S. Chen and Pengtao Xie. MedDialog: Large-scale Medical Dialogue Datasets. Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020.

We are working on Collaborative or Cooperative Autonomous driving and sensing focusing on efficient processing (in terms of usage of data and compute) of image and videos for assisted and autonomous driving applications. The project is attempting to build deep learning algorithms which can effectively combine data from other autonomous systems for safety applications. We are trying to combine data from all the sensors on other cars for self-driving cars for safety applications. The sensors include LiDAR, Camera and radars, CAN data and so on.

Publications: [1] Bansal K, Rungta K, Bharadia D. RadSegNet: A Reliable Approach to Radar Camera Fusion. arXiv preprint arXiv:2208.03849. 2022 Aug 8. [2] Dureppagari HK, Dinesha U, Wu R, Ganji S, Ko WH, Shakkottai S, Bharadia D. Realtime intelligent control for NextG cellular radio access networks. InProceedings of the 20th Annual International Conference on Mobile Systems, Applications and Services 2022 Jun 27 (pp. 567-568). [3] Givehchian H, Bhaskar N, Herrera ER, Soto HR, Dameff C, Bharadia D, Schulman A. Evaluating Physical-Layer BLE Location Tracking Attacks on Mobile Devices. In2022 IEEE Symposium on Security and Privacy (SP) 2022 May 22 (pp. 1690-1704). IEEE. [4] Arun A, Ayyalasomayajula R, Hunter W, Bharadia D. P2SLAM: Bearing Based WiFi SLAM for Indoor Robots. IEEE Robotics and Automation Letters. 2022 Jan 25;7(2):3326-33. [5] Wu Y, Ayyalasomayajula R, Bianco MJ, Bharadia D, Gerstoft P. Sound source localization based on multi-task learning and image translation network. The Journal of the Acoustical Society of America. 2021 Nov 5;150(5):3374-86. [6] Zhao M, Chang T, Arun A, Ayyalasomayajula R, Zhang C, Bharadia D. ULoc: Low-power, scalable and cm-accurate UWB-tag localization and tracking for indoor applications. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies. 2021 Sep 14;5(3):1-31. [7] Wu Y, Ayyalasomayajula R, Bianco MJ, Bharadia D, Gerstoft P. Sslide: Sound source localization for indoors based on deep learning. InICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2021 Jun 6 (pp. 4680-4684). IEEE.

This namespace is dedicated to the development and validation of a novel ECG parameterization tool. The project focuses on achieving accurate and stable analysis of ECG waveform shapes across multiple datasets consisting of healthy individuals. Key objectives include benchmarking the tool against established datasets to validate the stability of waveform shape parameters over time and examining their relationships with demographic covariates such as age, sex, and BMI. This work is critical for advancing the reliability and applicability of ECG parameterization in clinical and research settings.

The Ecological Forecasting Initiative is an international grassroots consortium aimed at building and supporting an interdisciplinary community of practice around near-term (daily to decadal) ecological forecasts. https://ecoforecast.org

We are an interdisciplinary team that develops and applies forecasts of ecosystem dynamics. We are an action-oriented research center: we collect and analyze environmental data; build and share ecological models and software; create and assess a diversity of forecasting methods; translate and communicate forecasts for decision support; and partner and engage with forecast users. We strive to advance the discipline of ecosystem forecasting globally by leading cutting-edge forecasting research, education, and community engagement.

Publications: https://doi.org/10.1002/fee.2616

This namespace will facilitate collaboration for the Schmidt Sciences Oxford Research Software Engineering program as part of the Eric & Wendy Schmidt AI in Science Postdoctoral Research Fellowship program.

Public EDEX server for weather data

Edgeslab is a research lab run by Elena Zheleva in the Computer Science department at UIC. The research topics are primarily concerned with data science and relational learning.

Publications: https://www.cs.uic.edu/~elena/#lab

Our research will be focused on ML-algorithms in the context of education and documentation.

This namespace is dedicated to USRA's Earth from Space Institute (EfSI) for compute resources and research.

Publications: N/A

1. Automatic web-based workflow for structural mutation of protein residues based on total charge. 2. Machine learning techniques for automatic protein charge modulation, and structural verification of mutated protein structures and activities using Molecular Dynamics. 3. Machine learning techniques for gene editing and functional analysis of CRISPR/Cas9 and CRISPR/Cas12a proteins. 4. High performance and distributed computing applications using systems connected by optical networks. 5. GUI desktop containerization accelerated with NVIDIA GPUs.

Publications: https://github.com/selkies-project/docker-nvidia-glx-desktop https://github.com/selkies-project/docker-nvidia-egl-desktop

Official elasticsearch

Elastiflow, sflow, InMon Traffic Sentinel deployment

Educational activities related to showing students high availability services

This is part of NSF MRI associated with m3learning namespace

Educational ENGR131 for the fall small class for introduction to python

Expanse federated namespace for testing the federation layer with toy workflows

Publications: “Towards a Dynamic Composability Approach for using Heterogeneous Systems in Remote Sensing”, Ilkay Altintas, Ismael Perez, Dmitry Mishin, Adrien Trouillaud, Christopher Irving, John Graham, Mahidhar Tatineni, Thomas DeFanti, Shawn Strande, Larry Smarr, Michael L. Norman. Proceedings of the IEEE 18th International Conference on e-Science, Salt Lake City, Utah, USA, October 11-14 2022. (Accepted)

Jupyter Lab environment for environmental data science resources for Universities Space Research Association (USRA)

This Namespace is dedicated to USRA's Environmental Analytics Group focusing on applications of Machine Learning in a variety of Earth science research such as wildfire, air quality, earthquake, floods, etc.

Envoy gateway for AI built on top of Envoy Proxy. Used for intelligent AI routing.

Aimed at making it easy to adopt, use, and manage Envoy Proxy. Deploy as a Standalone or Kubernetes-based API Gateway, implementing and extending the Kubernetes Gateway API.

Our work focuses on learning representations of robot motion and perception capabilities as well as of the environment the robots are operating in. We are interested in system identification from trajectory data for robot modeling, 3D scene reconstruction from RGBD and LiDAR measurements, model-based reinforcement learning, and vision-language models for robot task planning and execution.

Publications: https://existentialrobotics.org/pages/publications.html

Our work focuses on learning representations of robot motion and perception capabilities as well as of the environment the robots are operating in. We are interested in system identification from trajectory data for robot modeling, 3D scene reconstruction from RGBD and LiDAR measurements, model-based reinforcement learning, and vision-language models for robot task planning and execution.

Publications: https://existentialrobotics.org/pages/publications.html

Building AI-enabled data visualization interfaces: NAIRR Classroom

Testbed for Experimental Social Science Lab migration.

ESSRN is an initiative of UCs Berkeley, Davis, Irvine, and Santa Barbara and of the National Research Platform to build a network of social science laboratories serving as a national and international hub for experimental social science research. ESSRN will seamlessly accommodate lab, field, and online experiments on a vast and diverse multi-institutional subject pool, blurring the lines between the traditional lab and online recruitment platforms.

Etherpad system namespace provides etherpad installation - the collaborative markdown text editor.

EVL K8s research and application development towards the optiputer project and beyond. SAGE3 application and general UIC ML applications.

Publications: L. Long, T. Bargo, L. Renambot, M. Brown and A. E. Johnson, "Composable Infrastructures for an Academic Research Environment: Lessons Learned," 2022 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), Lyon, France, 2022, pp. 1209-1214, doi: 10.1109/IPDPSW55747.2022.00208. Composable Infrastructures for an Academic Research Environment: Lessons Learned, First Workshop on Composable Systems, COMPSYS 2022 Moving from Composable to Programmable, First Workshop on Composable Systems, COMPSYS 2022 PEARC22 - TITLE: CHI-in-a-Box: Reducing Operational Costs of Research Testbeds AUTHORS: Kate Keahey, Jason Anderson, Michael Sherman, Cody Hammock, Zhuo Zhen, Jenett Tillotson, Timothy Bargo, Lance Long, Taimoor Ul Islam, Sarath Babu and François Halbach Nurit Kirshenbaum, Kylie Davidson, Jesse Harden, Chris North, Dylan Kobayashi, Ryan Theriot, Roderick S. Tabalba, Michael L. Rogers, Mahdi Belcaid, Andrew T. Burks, Krishna N. Bharadwaj, Luc Renambot, Andrew E. Johnson, Lance Long, and Jason Leigh. 2021. Traces of Time through Space: Advantages of Creating Complex Canvases in Collaborative Meetings. Proc. ACM Hum.-Comput. Interact. 5, ISS, Article 502 (November 2021), 20 pages. DOI:https://doi.org/10.1145/3488552 K. Bharadwaj et al., "Securing Collaborative Work in Wide-band Display Environments," 2021 IEEE 7th International Conference on Collaboration and Internet Computing (CIC), 2021, pp. 26-34, doi: https://doi.org/10.1109/CIC52973.2021.00014. M. Ragonea, M. T. Saraya, L. Long, R. Shahbazian-Yassara, F. Mashayeka, V. Yurkiv, "Deep learning for mapping element distribution of high-entropy alloys in scanning transmission electron microscopy images,” Computational Materials Science, Volume 201, January 2022, 110905, https://doi.org/10.1016/j.commatsci.2021.110905.

This project focus on implementation of charged particle tracking pipeline as a Triton Inference Server. Clients implemented in ACTS will send track-finding requests to the Triton server and the server will return track candidates to the client after processing. The pipeline contains several track reconstruction algorithms. Because of the heterogeneity and dependency chain of the pipeline, we will explore different server settings to maximize the throughput of the pipeline, and we will study the scalability of the inference server and time reduction of the client.

working with eye-tracking and human attention. the project will deal with how human attention affects different deep learning models.

Colleting eye tracking data for image captioning and analysing its role in captioning

System namespace - for falco security alerting setup on 5/6/24

Namespace where rasa-bot for rocket chat will be implemented .

Protein folding via Folding@Home for COVID-19

Fusion plasma simulation, mostly using CGYRO. Relies on many-node GPU resources.

Generative AI research for multimodal synthetic medical data generation

Publications: https://scholar.google.com/citations?user=gE-WTF8AAAAJ&hl=en

Building LLMs for agents modeling, on multi-gpu clusters.

Gas detection from spectral images for remote sensing applications

Gatekeeper is a validating and mutating webhook that enforces CRD-based policies executed by Open Policy Agent, a policy engine for Cloud Native environments hosted by CNCF as a graduated project.

Generative AI lab space for Universities Space Research Association.

Our lab works on generating interpretable explanations from opaque ML models and using them to make more robust decisions. We use Nautilus for training/running DNNs, rules lists, and generative models to generate a novel dataset of failure cases for analysis.

Publications: None.

GitLab deployment

gnmic-dev desktop environment using containerlab and noVNC

JupyterHub instance for GP-ENGINE related compute

Research for Dr. Jordan Malof's lab at the University of Missouri

Deep learning research using remote sensing data for post-wildfire assessment

Research and teaching for Mizzou research lab

This namespace is going to be used to teach data science and analytics students how to use cloud computing to perform big data related task and computing

Pods and jobs for the classification and detection of Boma settlements

Experiments for pretraining DNN models

Deployment of Apache Airflow and MLFlow. Testing MLOps architecture for EO imagery processing and archiving

Research for the MINDFUL lab at the University of Missouri

Research for MU lab

Test Windows Virtual Machine with KubeVirt

Publications: None

Research for Lab at MU

Research for lab at MS&T

Build machine learning and deep learning models to evaluate the development of some diseases and also patient outcomes.

The research in this namespace research focuses on biomedical informatics, explainable AI, quantum computing, cybersecurity, and spatial Big Data analytics

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Running JupyterHub for research related to quantum computing in University of Missouri

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Our project leverages Artificial Intelligence, Big data analytics, and Machine learning to revolutionize transportation systems to enhance safety, resilience and efficiency.

Publications: IEEE Intelligent Transportation Systems, Transporation Research, IEEE/Computer Vision and Pattern Recognition

The Research Informatics Lab (RILab) at the University of Missouri is a hub of innovation in health informatics, focusing on enhancing patient care through advanced data analysis and informatics technologies. The lab's research encompasses developing support tools for complex healthcare decisions and pioneering patient data privacy and management methods. Significant contributions include predictive modeling for managing chronic diseases and creating tools to improve treatment outcomes.

Research JupyterHub instance for resarch under the GP-ENGINE grant

Pods and Jobs for the SDSS Tutorial 2025 on Accelerating Data Science Workflows with Kubernetes

Namespace for developing training material for GP-ENGINE and Nautilus

namespace for people participating in gp-engine tutorials related to Kubernetes and Nautilus.

Namespace for Jupyter front end for gp-engine tutorials

Software engineering research projects; Deep learning models for software languages

Publications: https://scholar.google.com

Software engineering research projects; Deep learning models for software languages

Publications: https://scholar.google.com

Resources for UNT research and teaching

JupyterHub demonstrations for institution deployment

JupyterLab instance for CS education and experimentation with Binder

Research and teaching for Mizzou lab

Research and teaching for Mizzou lab

JupyterHub for Mizzou CS Dept

Research on remote sensing and health informatics to determine disease patterns

Publications: "None

Mizzou DSA Parallel Computing Course

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Research and teaching for Mizzou eecs department

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Mizzou DSA Parallel Computing Course

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Namespace for JupyterHub running for Mizzou HPC.

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Research for Mizzou lab

Publications:

Research for MU grad student

Publications: None

Namespace to be used for research and development of Kubernetes workshop

Research and teaching for Mizzou eecs department

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Research and teaching for Mizzou lab

Research and teaching for Mizzou lab

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Research and teaching for Mizzou lab

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Research and teaching for Mizzou lab

Modeling visual reasoning processes to create the best model there is

Research and teaching for graduate student in HPDI Lab at the University of Missouri

Research and teaching for Mizzou lab

Research and Development namespace for Research Computing Support Services in the Division of IT at the University of Missouri

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Research and teaching for Mizzou ViGiR lab

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

Namespace for Dr. DeSouza's lab from the University of Missouri

Publications: None

JupyterHub environment for OU Machine Learning Practice and OU Deep Learning Practice classes.

Various testing efforts and socialization of the platform.

AI/ML GPU benchmarks for Tom DeFanti using NVIDIA MLCommons benchmarking suite. Specifically focusing on comparing A100 and our 4090 nodes.

Use ML (primarily graph ML and NLP) methods to extract and analyze information from unstructured text data.

Apache guacamole deployment - the VNC/RDP/Console client to use for remote IPMI management and accessing remote GUI apps

Gary's Unbelievable Research Unit (GURU) is carrying out a wide range of projects using deep learning. Recent projects include using deep learning to recognize speech and perform diarization in This American Life episodes, using deep learning to segment mouse cardiac MRI to speed animal research in heart disease in collaboration with the UCSD Medical School, mapping from small molecule NMR to a cluster space where similar molecular structures are near one another in the space to speed structure elucidation for natural products in collaboration with researchers at Scripps Institution of Oceanography, using GANS to modify faces to alter their first impressions, developing methods to speed up convergence of deep learning, and projects in computational cognitive neuroscience, modeling the human visual system using anatomical constraints in order to explain how the visual system works.

Publications: Henry Huanru Mao, Shuyang Li, Julian McAuley, Garrison W. Cottrell (2020) Speech Recognition and Multi-Speaker Diarization of Long Conversations. arXiv preprint arXiv:2005.08072. Hammad A. Ayyubi, Md. Mehrab Tanjim, Julian J. McAuley, Garrison W. Cottrell (2020) Generating Rationales in Visual Question Answering. arXiv preprint arXiv:2004.02032. Thomas Bachlechner, Bodhisattwa Prasad Majumder, Huanru Henry Mao, Garrison W Cottrell, Julian McAuley (2020) ReZero is all you need: Fast convergence at large depth. arXiv:2003.04887. Raphael Reher, Hyun Woo Kim, Chen Zhang, Huanru Henry Mao, Mingxun Wang, Louis-Félix Nothias, Andres Mauricio Caraballo-Rodriguez, Evgenia Glukhov, Bahar Teke, Tiago Leao, Kelsey L Alexander, Brendan M Duggan, Ezra L Van Everbroeck, Pieter C Dorrestein, Garrison W Cottrell, William H Gerwick (2020). A Convolutional Neural Network-Based Approach for the Rapid Characterization of Molecularly Diverse Natural Products. Journal of the American Chemical Society 142(9):4114-4120. (supplementary material). Yueying Li, Hao-Bing Yu, Yi Zhang, Tiago Leao, Evgenia Glukhov, Marsha L Pierce, Chen Zhang, Hyunwoo Kim, Huanru Henry Mao, Fang Fang, Garrison W Cottrell, Thomas F Murray, Lena Gerwick, Huashi Guan, William H Gerwick (2020) Pagoamide A, a Cyclic Depsipeptide Isolated from a Cultured Marine Chlorophyte, Derbesia sp., Using MS/MS-Based Molecular Networking. Journal of Natural Products 83(3):617-625. Mao, Huanru Henry, Majumder, Bodhisattwa Prasad, McAuley, Julian and Cottrell, Garrison W. (2019) Improving Neural Story Generation by Targeted Common Sense Grounding. In Empirical Methods in Natural Language Processing (EMNLP-19). Attala, Chad, Song, Amanda, Tam, Bartholomew, Rathis, Asmitha, and Cottrell, Garrison W. (2019) Modifying social dimensions of faces with ModifAE. In A.K. Goel, C.M. Seifert, & C. Freksa (Eds.), Proceedings of the 41st Annual Conference of the Cognitive Science Society (pp. 105-111). Montreal, QB: Cognitive Science Society. Donahue, Chris, Mao, Huanru Henry, Li, Yiting Ethan, Cottrell, Garrison W., and McAuley, Julian (2019) LakhNES: Improving multi-instrumental music generation with cross-domain pre-training. In Proceedings of the International Society for Music Information Retrieval Conference (ISMIR-19).

Research lab that studies the formation, lives, and explosive deaths of stars across cosmic time from their fossils as black holes and neutron stars.

HaoSu Image Service: Frontend for HaoSu S3. Hosts scalable image processing services on HaoSu nodes.

HAProxy system install - the ingress controller providing SSL termination and proxying http requests

SENSE FE RM for OpenNSA AutoGOLE

HedgeDoc lets you create real-time collaborative markdown notes.

Hengenglab is a Neuroscience lab lead by Kieth Hengen out of Washington University in St. Louis.

hgx-a100-u55c novnc development environment node-2-1

We observed a large number of failed transfers in science communities. Our hypothesis is some of these errors might be going past TCP checksum. These errors can be caused by many things - software bugs, hardware faults, and others. We have a prototype client/server that adds error checking headers to catch any error going past TCP checksum. The way the client/server works is this: we give the server a set of files. The clients transfer this set of files over and over again. It logs any errors it sees and throws away the rest. We then analyze the errors offline. We will create multiples pods - one with the server and others with the clients. All pods will utilize files in persistent storage. We do not anticipate large amounts of data to/from outside NRP.

Publications: https://ieeexplore.ieee.org/abstract/document/10206520

Used for horizontal scalability of HLS4ML model codesign

High Performance Computing namespace for parallel processing tasks using Dask

Kubernetes environment for the HSRN team. Mainly used for testing.

Experiment with using BGPalerter to monitor Internet2 Member Resources

Testing multi-cluster integration and performance for research computing and performance across various national resources (commercial cloud, Jetstream2, FABRIC, etc.)

Publications: https://github.com/MiddelkoopT/nautilus-tutorial

Danswer AI Chatbot. Adding some more characters to meet the arbitrary requirement.

Internet2 Network Engineering name space to perform various network related activities.

A simple webtool to perform RPKI Validation of prefixes.

Temporary namespace to demonstrate concepts at the Internet2 Technology Exchange

Sandbox for exploring the National Research Platform to develop an understanding and materials for the Research Computing and Data (RCD) community.

ML training in support of the IceCube project, plus used as a development namespace.

Publications: https://icecube.wisc.edu/science/publications/

Elasticsearch on igrok nodes - cluster monitoring and logs collection

How to represent an image? While the visual world is presented in a continuous manner, machines store and see the images in a discrete way with 2D arrays of pixels. In this paper, we seek to learn a continuous representation for images. Inspired by the recent progress in 3D reconstruction with implicit function, we propose Local Implicit Image Function (LIIF), which takes an image coordinate and the 2D deep features around the coordinate as inputs, predicts the RGB value at a given coordinate as an output. Since the coordinates are continuous, LIIF can be presented in an arbitrary resolution. To generate the continuous representation for pixel-based images, we train an encoder and LIIF representation via a self-supervised task with super-resolution. The learned continuous representation can be presented in arbitrary resolution even extrapolate to ×30 higher resolution, where the training tasks are not provided. We further show that LIIF representation builds a bridge between discrete and continuous representation in 2D, it naturally supports the learning tasks with size-varied image ground-truths and significantly outperforms the method with resizing the ground-truths.

Publications: Yinbo Chen, Sifei Liu, Xiaolong Wang. Learning Continuous Image Representation with Local Implicit Image Function. CVPR 2021. Yinbo Chen, Zhuang Liu, Huijuan Xu, Trevor Darrell, Xiaolong Wang. Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learning. International Conference on Computer Vision (ICCV), 2021.

Photo hosting with AI and auto recognition. Uses ceph for storage.

Namespace for developing and testing course materials for AI/ML courses in EECS at UC Berkeley.

Publications: https://inst.eecs.berkeley.edu

SG-based dynamic evaluation of MLLMs. First, SG is constructed given an image. Then dynamic question is asked based on SG. Finally SG-based edit can be used to edit the image to evaluate the model.

Isaac Sim service deployments NVIDIA Isaac Sim™ is a reference application built on NVIDIA Omniverse that enables developers to simulate and test AI-driven robotics solutions in physically based virtual environments.

Playground namespace, for testing purposes. Belongs to Igor Sfiligoi.

The UCSC Private Nautilus Namespace is a dedicated environment for undertaking projects that require intensive computational work. This namespace provides access to high-performance computing resources that can handle complex and computationally demanding tasks. The namespace is designed to support projects with specific requirements, such as those that demand very high levels of hardware performance. In this namespace, users can undertake projects with large data sets, complex simulations, and intricate algorithms. The namespace provides a secure and isolated environment for users to perform their computations, ensuring data privacy and security. Overall, the UCSC Private Nautilus Namespace is a valuable resource for researchers and scientists looking to tackle large-scale projects that require significant computational resources. It provides a powerful and flexible environment that can accommodate a wide range of research activities, making it an essential tool for researchers at UCSC.

Temporary tests of concepts associated with leveraging k8s environments for analysis to assist with research engagement.

Publications: NA

WEB conference software Jitsi for videoconferencing

JJGRAHAM Namespace for development foo and such and so on

Sleep is a crucial part of the daily activity patterns of mammals. However, in marine species that spend months or entire lifetimes at sea, the location, timing, and duration of sleep may be constrained. To understand how marine mammals satisfy their daily sleep requirements while at sea, we monitored electroencephalographic activity in wild northern elephant seals (Mirounga angustirostris) diving in Monterey Bay, California. Brain-wave patterns showed that seals took short (less than 20 minutes) naps while diving (maximum depth 377 meters; 104 sleeping dives). Linking these patterns to accelerometry and the time-depth profiles of 334 free-ranging seals (514,406 sleeping dives) revealed a North Pacific sleepscape in which seals averaged only 2 hours of sleep per day for 7 months, rivaling the record for the least sleep among all mammals, which is currently held by the African elephant (about 2 hours per day).

Publications: Kendall-Bar, Jessica M., Terrie M. Williams, Ritika Mukherji, Daniel A. Lozano, Julie K. Pitman, Rachel R. Holser, Theresa Keates et al. "Brain activity of diving seals reveals short sleep cycles at depth." Science 380, no. 6642 (2023): 260-265.

Deep learning for AI in Education: We are working on core natural language processing (NLP) technologies for an AI partner that interacts with students and teachers and helps them work and learn together more effectively. This is work being carried out as part of the NSF AI Institute for Student-AI Teaming. Robust semantic understanding algorithms will enable the partner to interact with students and teachers in a more natural and dynamic manner. We are developing deep learning methods for the AI partner to analyze the conversation topic and lesson plan materials, and participate in classroom discussions in a manner that promotes engagement and critical thinking from a diverse background of students. We are also working on incorporating semantic information (Abstract Meaning Representation graphs) into machine translation, and temporal language modeling for predicting future paper abstracts.

Publications: Improving Neural Machine Translation with the Abstract Meaning Representation by Combining Graph and Sequence Transformers. Changmao Li and Jeffrey Flanigan. NAACL 2022 Workshop on Deep Learning on Graphs for Natural Language Processing. Google Scholar: https://scholar.google.com/citations?user=XpIsORcAAAAJ&hl=en

Jobs admission controller (system namespace) - remembers who ran which job

Baseball Statistics Project Objective: Analyze historical player statistics across the league, break the data down per team, and compare with farm league players. Data Sources: • Historical Player Statistics: Publicly available datasets from sports databases or APIs. • Team Data: Data from individual team websites or sports analytics platforms. • Farm League Data: Data from minor league databases or sports organizations. Data Stores: • PostgreSQL: Store structured data such as player statistics, team records, and game results. • Graph Database (e.g., Neo4j): Model relationships between players, teams, and leagues. • Data Lake (e.g., Rook S3): Store raw data files, including CSVs, JSONs, and other formats. Key Features: • Data Aggregation: Combine data from different leagues and teams. • Comparative Analysis: Compare player performance across different levels and teams. • Predictive Modeling: Use machine learning to predict player progression from farm leagues to major leagues. Potential Challenges: • Data Volume: Handle large volumes of historical data. • Data Consistency: Ensure data from different sources is consistent and comparable. Next Steps • Define Scope: Clearly outline the scope and objectives of your project. • Gather Data: Identify and collect the necessary datasets. • Design Schema: Plan the database schema for each data store. • Develop: Implement the project using the chosen technologies. • Test and Validate: Ensure the accuracy and reliability of your analysis.

Juice remote GPU over IP test deployment bla bla bla

Jump service NRP Nautilus remote access ............

System namespace, Jupyterlab deployment. Used by multiple users for various tasks.

Publications: http://ucsd-prp.gitlab.io/userdocs/running/jupyter/ No pubs

Eastern jupyterlab system namespace. Used by multiple users for variuos tasks.

This is a demo namespace used for testing, training, and outreach for the Nautilus system. Users in this namespace are stepping through tutorials and exploring the capabilities of the system.

Publications: None.

In this project, we are developing an optimized data streaming pipeline for real-time graph data analysis, using a community detection algorithm as the first test application. In this system, streaming data sets will be fed to the pipeline where a preprocessing phase will be performed to clean the dataset. In this initial twitter-based test case, the preprocessing step will also include fetching the co-occurrent hashtags from tweets. The resulting data is then partitioned and sent to a Kafka cluster by multiple producers using a stream-specific partitioning mechanism, generally based on a hash of the data being analyzed (e.g. the hashtags). Consumers of Kafka topics are responsible for reading the data from the topics and constructing rows of the adjacency matrix for the graph being analyzed. Each portion of this adjacency matrix will then be saved to be accessible by relevant graph analysis algorithms, e.g. PASCAL-G algorithm in the case of Twitter community analysis. PASCAL-G is a probabilistic stream clustering analysis on graphs that needs the adjacency matrix as the input. A vital purpose of this research is to increase the elasticity of this pipeline and to make it scalable. To do so, we are researching how to best optimize the performance of this overall pipeline in terms of latency and throughput, by adjusting Kafka server parameters such as the number of partitions, the size of each partition, the number of producers and consumers, the acknowledgment policy, etc. This pipeline uses a containerized orchestration in which the zookeeper, Kafka, producers, consumers, and the machine learning algorithm is containerized in a docker container. This pipeline has been already developed and debugged on a local server using Docker Compose. We are immigrating this pipeline to Kubernetes on Nautilus to test the pipeline’s scaling and policies for controlling its parameters, creating and analyzing the distribution of producer, Kafka, and consumer’s latency and throughput, and modeling and optimizing this distribution.

Foundational AI and machine learning: Aligned with USD's AI programs, this is a place where everyone—regardless of background—can thrive. Our passion lies in striving for excellence, driving AI innovation, and supporting one another in the pursuit of success.

Publications: https://www.ai-research-lab.org

Keda is a workload scaler with extended features compared to HPA

Yum repo and kernel builds - provides yum repo to centos nodes

ML-based microbiome research for the KnightLab students and faculty.

Active learning platform using pytorch and python. Unpublished

Kansas State University utilizes NRP resources primarily to run LLMs for Natural Language Processing. Additionally students utilize resources for other NLP tasks such as; Triple Extraction and Stance Detection.

JupyterHub instance for miscellaneous research/teaching workflows

Publications: None.

System namespace for Nautilus cluster

Kubeflow install

Kubevirt installation - provides ability to run VMs in k8s

Deep learning models in genomics: Exploring new deep learning architectures to improve the classification accuracy of deep learning models in genomics. More generally, work focuses on leveraging deep learning for genomics in conjunction with interpretation techniques to extract novel insights about regulatory genomics. Decoding regulatory DNA sequence in keratinocyte differentiation: Development and differentiation are biological processes that involve cascades of transcription factors interacting with dynamic chromatin landscapes to produce cell-type specific transcriptional programs. Epidermal differentiation, in which a self-renewing progenitor keratinocyte becomes a terminally differentiated keratinocyte, is well suited for studying fine-grained changes in chromatin and transcription and addressing fundamental questions about the dynamic combinatorial logic of regulation. To answer these questions, genomic profiling of transcriptional state (using 3' RNA-seq) and chromatin state (using ATAC-seq and ChIP-seq on histone marks) was captured at 12 hour intervals across 6 days of in vitro differentiation of primary keratinocytes. We inferred transcriptional and epigenetic trajectories across time to elucidate dynamically coordinated modules of genes and regulatory elements. We then developed deep, multi-task convolutional neural networks to learn predictive DNA sequence drivers of chromatin dynamics. To discover motifs and coordinated motif sets (grammars) from the neural net, we used backpropagation methods to derive nucleotide level importance scores in regulatory elements across time that are then used to extract grammars that are predictive of accessibility. We use these grammars in conjunction with expression and chromosome conformation assays to annotate functional modules that define known and novel differentiation programs. The resulting framework provides a generalizable approach to dissecting dynamic maps of combinatorial regulation encoded in DNA sequence.

Publications: Peer-reviewed journal publications, conference proceedings EM with Bias-Corrected Calibration is Hard-To-Beat at Label Shift Adaptation Alexandari A, Kundaje A*, Shrikumar A* arXiv preprint arXiv:1901.06852, 2019 Jan 21 (Accepted to ICML 2020) Integrating regulatory DNA sequence and gene expression to predict genome-wide chromatin accessibility across cellular contexts Nair S, Kim DS, Perricone J, Kundaje A Bioinformatics, Volume 35, Issue 14, July 2019, Pages i108–i116, https://doi.org/10.1093/bioinformatics/btz352 (PMID: 31510655) (Proceedings of ISMB 2019) The Kipoi repository accelerates community exchange and reuse of predictive models for genomics Avsec Ž, Kreuzhuber R, Israeli J, Xu N, Cheng J, Shrikumar A, Banerjee A, Kim DS, Beier T, Urban L, Kundaje A*, Stegle O*, Gagneur J* Nat Biotechnol. 2019 May 28 DOI: 10.1038/s41587-019-0140-0 (PMID: 31138913) Kipoi: accelerating the community exchange and reuse of predictive models for regulatory genomics Avsec Z, Kreuzhuber R, Israeli J, Cheng J, Urban L, Banerjee A, Xu N, Shrikumar A, Ouwehand WH, Kundaje A*, Stegle O*, Gagneur J* ICML 2018 Workshop for Computational Biology BPNet: Learning single-nucleotide resolution predictive models of in vivo transcription factor binding from ChIP-nexus data Avsec Z, Israeli J, Fropf R, Weilert M, Zeitlinger J, Kundaje A ICML 2018 Workshop for Computational Biology Preprints: Fourier-transform-based attribution priors improve the interpretability and stability of deep learning models for genomics Tseng AM, Shrikumar A, Kundaje A bioRxiv 2020.06.11.147272; doi: https://doi.org/10.1101/2020.06.11.147272 Single-cell epigenomic identification of inherited risk loci in Alzheimer's and Parkinson's disease Corces MR, Shcherbina A, Kundu S, Gloudemans MJ, Fresard L, Granja JM, Louie BH, Shams S, Bagdatli ST, Mumbach MR, Liu B, Montine KS, Greenleaf WJ, Kundaje A, Montgomery SB, Chang HY, Montine TJ bioRxiv 2020.01.06.896159; doi: https://doi.org/10.1101/2020.01.06.896159 Learning cis-regulatory principles of ADAR-based RNA editing from CRISPR-mediated mutagenesis Liu X, Sun T, Shcherbina A, Li Q, Kappel K, Jarmoskaite I, Ramaswami G, Das R, Kundaje A*, Li JB* bioRxiv 840884; doi: https://doi.org/10.1101/840884 A genome-wide almanac of co-essential modules assigns function to uncharacterized genes Wainberg M, Kamber RA, Balsubramani A, Meyers RM, Sinnott-Armstrong N, Hornburg D, Jiang L, Chan J, Jian R, Gu M, Shcherbina A, Dubreuil MM, Spees K, Snyder MP, Kundaje A*, Bassik MC* bioRxiv 827071; doi: https://doi.org/10.1101/827071 Deep learning at base-resolution reveals motif syntax of the cis-regulatory code Avsec Ž, Weilert M, Shrikumar A, Alexandari A, Krueger S, Dalal K, Fropf R, McAnany C, Gagneur J, Kundaje A*, Zeitlinger J* bioRxiv 737981; doi: https://doi.org/10.1101/737981 TF-MoDISco v0.4.4.2-alpha: Technical Note Shrikumar A, Tian K, Shcherbina A, Avsec Z, Banerjee A, Sharmin M, Nair S, Kundaje A ArXiv e-prints:1811.00416, 2018 Nov 1 A Flexible and Adaptive Framework for Abstention Under Class Imbalance Shrikumar A, Alexandari A, Kundaje A ArXiv e-prints [Internet]. 2018 Feb 20

UCSD's Learning, meaning, and natural language lab is run by PI Alex Warstadt. The group focuses on interdisciplinary research in linguistics, computational cognitive modeling, and natural language processing. We use advances in machine learning to understand why human language is the way it is, how children come to acquire it, and how information is conveyed across multiple channels. We use insights from linguistics and cognitive science to advance data-efficient learning in LMs and interpret how LMs learn and represent grammatical structures and meaning.

Librareome is an ongoing project that uses tools built by the ARNO team to create a UCSD Campus Scale AR SYSTEM inspired by Vernor Vinge's quintessential novel RAINBOWS END.

Publications: None

LIGO namespace for running Rucio for management and transfers of instrumental data between the LIGO observatories in Hanford WA and Livingston LA, and to central archives at Caltech

Sandboxed automated testing of emfollow as part of LLAI

Our project focuses on conducting a rigorous evaluation of the security aspects pertaining to Large Language Models (LLMs). Through meticulous validation, thorough testing, and in-depth security assessments, we aim to enhance the understanding and fortification of LLMs against potential vulnerabilities and threats. This endeavor encompasses a systematic exploration of various dimensions of LLM security, with the ultimate goal of fostering robustness, reliability, and trustworthiness in these pivotal technologies.

Current deep learning models rely on two assumptions, 1) consistent distribution between train and deployment data and 2) balanced class distributions. In practice, however, these assumptions are frequently violated and model performance can degrade significantly. In this project, we propose to break these constraints by developing algorithms that are jointly robust to domain-shift and long-tailed distributions. This is essential to guarantee robust and label-efficient visual recognition performance on a variety of applications, including scene understanding, autonomous vehicles, robotics, augmented reality, AI on edge devices, and virtual assistants.

Namespace for CRISPS cell centric image similarity search and MRI project

Maddy mail server - used for sending all users notifications

Testing and benchmarking for NRP. NVIDIA GPUs and Xilinx FPGAs

Land data products such as fuel maps and land cover maps are critical for many applications including land use analysis, bio-diversity conservation, and wildfire management. Current products provide essential land data and are widely used by various agencies across the nation. However, these products are generated very infrequently (e.g., every few years) and based on medium-resolution imagery that do not provide the granularity possible with high-resolution imagery. Our research proposes an approach to generate products as needed, based on up-to-date imagery, and at scale. Specifically, our research goals are to generate more frequent products by creating maps as needed from satellite imagery, more accurate products by using up-to-date, high-resolution satellite imagery, and scalable products by utilizing machine learning to automate the process. The approach we use extracts features from satellite images using a deep learning model. The resulting feature vectors are then used for classifying or segmenting the images in order to generate land cover maps. We plan to extend our work to multi-spectral imagery, larger and more varied regions, and other types of land data products.

Publications: Nguyen, M. H., Block, J., Crawl, D., Siu, V., Bhatnagar, A., Rodriguez, F., Kwan, A., Baru, N., and Altintas, I. “Land Cover Classification at the Wildland Urban Interface using High-Resolution Satellite Imagery and Deep Learning,” in the 2018 IEEE International Conference on Big Data

This project investigates techniques to perform automatic left ventricle (LV) segmentation and volume estimation from cardiac magnetic resonance imaging (MRI). The LV is the largest chamber in the heart and plays a critical role in cardiac function. Cardiac imaging such as MRI provides an non-invasive way to study cardiac structure and function, and is an invaluable tool in heart disease diagnosis. However, the process of analyzing cardiac images to perform LV segmentation is time-consuming, labor-intensive, and error-prone. Automating this process is thus essential in providing efficient and consistent analysis of cardiac images for diagnosing heart disease. We are applying deep and machine learning methods to create an analytics pipeline to automate this process. Methods are also examined for preprocessing cardiac images, performing semantic segmentation of the LV, as well as estimating LV volume.

Publications: M. H. Nguyen, E. Abdelmaguid, J. Huang, S. Kenchareddy, D. Singla, L. Wilke, M. Bobar, E. D. Carruth, D. Uys, I. Altintas, E. D. Muse, G. Quer, S. Steinhubl, “Analytics Pipeline for Left Ventricle Segmentation and Volume Estimation on Cardiac MRI using Deep Learning,” in Proc. IEEE 14th Int. Conf. on e-Science, 2018 E. Abdelmaguid, J. Huang, S. Kenchareddy, D. Singla, L. Wilke, M. H. Nguyen, and I. Altintas. "Left ventricle segmentation and volume estimation on cardiac mri using deep learning." arXiv preprint arXiv:1809.06247 (2018).

Matrix-synapse system namespace used for Matrix chat system. We build users community with it.

neural network models for animal exploration behavior

We use neural network models to understand human cognition, especially decision-making and planning.

We are from visual computing center. The cluster resources will be mainly used for vision and graphic research. The group is lead by Prof. Manmohan Chandraker at UCSD.

Publications: IROS 2020: Deep Keypoint-Based Camera Pose Estimation with Geometric Constraints (You-Yi, Rui, Hao, Manmohan) ECCV 2020: Single-Shot Neural Relighting and SVBRDF Estimation (Shen, Manmohan) ECCV 2020: Single-View Metrology in the Wild (Rui, Yannick, Federico, Jonathan, Kalyan, Manmohan) CVPR 2020: Inverse Rendering for Complex Indoor Scenes (Zhengqin, Mohammad, Kalyan, Ravi, Manmohan) CVPR 2020: Neural 3D Reconstruction of Transparent Shapes (Zhengqin, Yu-Ying, Manmohan) CVPR 2022: IRISformer: Dense Vision Transformers for Single-Image Inverse Rendering in Indoor Scenes (Rui Zhu, Zhengqin, Manmohan)

For large-scale data generation and rendering tasks, which could utilize dozens of nodes for days or weeks.

AI-Centric NextG Wireless: Developing Full Stack, Secure, Wireless Intelligence in Pursuit of the NextG

The Microelectronic Embedded Systems Laboratory is part of the Computer Science and Engineering department at UCSD and is led by Professor Rajesh Gupta. We are interested in most aspects of embedded computer systems and sensor networks. Our recent research focus has been on data mining, time-series, federated learnings, and large language models methodolgies particularly in sensing domain. Besides, Cyber-Physical Systems (especially smart buildings) and Internet of Things is another major direction that we've been working actively on.

MetalLB installation - provides external IPs to services

AgiSoft Metashape

A development namespace for Mohammad Sada, projects: Selkies, Coder, Seam, Omniverse, Vivado, Vitis....

Computing for projects in Gal Mishne's lab. Analysis of large-scale neural recordings and behavior

Revolutionary 4D microscopy creates unprecedented opportunities in cell biology and unprecedented datasets. Our mission is to be a pioneer in this new era, push forward, and lower the barriers to entry for the whole community in the process.

Publications: https://www.schoeneberglab.org/papers

mizzou test jupyterhub

Publications: Setting Up Research Jobs in Nautilus Kubernetes Hyper-Cluster: https://github.com/scottgs/Nautilus_Kubernetes_DeepLearning

We are leveraging Nautilus platform for teaching k8s related concepts and skills in different classrooms.

Publications: Setting up Kubernetes cluster for learning on: https://github.com/MizzouCloudDevOps/mizzou_nautilus_k8s_lab

Nautilus workspace for Machine Learning, Creative AI, Robotics, and Cultural Analytics experiments. We take imagination as a basis to explore, identify, measure and produce outputs of ML to extend and explore human imagination. Facilitating workshops, research, and courses through access to open source tools.

Publications: https://roberttwomey.com https://roberttwomey.com/writing/ https://go.unl.edu/aifilm https://dl.acm.org/doi/10.1145/3610978.3640577 https://cohab-lab.net

mlnx-p4

System namespace for monitoring stuff - grafana, prometheus, etc. Monitors everyting in cluster.

Kubeflow MPI operator - provides mpi workflows for kubernetes

Segmentation on Imagenet and multiscale attention model testing

Publications: https://scholar.google.com/citations?user=HKB-zg8AAAAJ&hl=en&oi=ao

I'm trying to run some chemistry computing models in the namespace

Publications: Not yet so far

Parallelized molecular docking based on the work of Daniel Medeiros et. al. (https://arxiv.org/abs/2410.10634). Using Apache Airflow as a workflow manager for computational chemistry pipelines. Jupyterhub as a serverless UI for computational chemistry as demonstrated by Christopher Woods et. al. (https://kccna18.sched.com/event/GraL).

JupyterHub for MTSU CSCI Software Stack (https://www.mtsu.edu/csc/ and https://github.com/Phillips-Lab-MTSU/CSCI-MTSU-JupyterHub)

Namespace for UCSD Astronomy & Astrophysics Assistant Professor Ethan Nadler's group

nautobot

UCSC NCG assign to Assel, for the knowledge distillation

The mission of NCMIR is to develop technologies to bridge understanding of biological systems between the gross anatomical and molecular scales and to make these technologies broadly available to biomedical researchers. NCMIR provides expertise, infrastructure, technological development, and an environment in which new information about the 3D ultrastructure of tissues, cells, and macromolecular complexes may be accurately and easily obtained and analyzed.

Publications: CDeep3M-Preview: Online segmentation using the deep neural network model zoo Matthias G Haberl, Willy Wong, Sean Penticoff, Jihyeon Je, Matthew Madany, Adrian Borchardt, Daniela Boassa, Steven T Peltier, Mark H Ellisman bioRxiv 2020.03.26.010660; doi: https://doi.org/10.1101/2020.03.26.010660

The scientific community generates vast amounts of data through research, experiments, and observations. Effective management of this data to support equitable discovery, access, analysis, communication, and sharing is critical for research, innovation, and science-driven decision making. Furthermore, broad and equitable access to diverse artificial intelligence (AI)-ready data repositories is crucial to realize the full potential of AI in responsibly advancing solutions to important scientific and societal problems at national and global scale. In response to these immediate needs, the National Data Platform (NDP) serves as a federated and extensible data ecosystem that fosters collaboration, innovation, and equitable data use, leveraging existing national Cyberinfrastructure capabilities. Connecting distributed computing and data CI systems to provide open access to data from disparate, often siloed repositories and other data sources necessitates a standardized process and services for ingestion, indexing, curation and data analysis. Through a ?removing-the-barriers?' approach to democratizing data, NDP combines needs assessment, co-design, and diversity-aware capacity building with ready-for-scale data CI capabilities, offering data and knowledge management services across the national CI ecosystem. As a national hub of interconnected data hubs, NDP facilitates data discovery and usage, drives responsible AI research and development, fosters scientific understanding, and supports decision-making, policy formation and societal impact. NDP focuses on several case studies in climate and related research areas to evolve data-centric workflows, including wildland fire, and natural and real-time hazards detection and decision making This effort is also supported by National Discovery Cloud for Climate (NDC-C) resources.

The scientific community generates vast amounts of data through research, experiments, and observations. Effective management of this data to support equitable discovery, access, analysis, communication, and sharing is critical for research, innovation, and science-driven decision making. Furthermore, broad and equitable access to diverse artificial intelligence (AI)-ready data repositories is crucial to realize the full potential of AI in responsibly advancing solutions to important scientific and societal problems at national and global scale. In response to these immediate needs, the National Data Platform (NDP) serves as a federated and extensible data ecosystem that fosters collaboration, innovation, and equitable data use, leveraging existing national Cyberinfrastructure capabilities. Connecting distributed computing and data CI systems to provide open access to data from disparate, often siloed repositories and other data sources necessitates a standardized process and services for ingestion, indexing, curation and data analysis. Through a ?removing-the-barriers?' approach to democratizing data, NDP combines needs assessment, co-design, and diversity-aware capacity building with ready-for-scale data CI capabilities, offering data and knowledge management services across the national CI ecosystem. As a national hub of interconnected data hubs, NDP facilitates data discovery and usage, drives responsible AI research and development, fosters scientific understanding, and supports decision-making, policy formation and societal impact. NDP focuses on several case studies in climate and related research areas to evolve data-centric workflows, including wildland fire, and natural and real-time hazards detection and decision making This effort is also supported by National Discovery Cloud for Climate (NDC-C) resources.

The scientific community generates vast amounts of data through research, experiments, and observations. Effective management of this data to support equitable discovery, access, analysis, communication, and sharing is critical for research, innovation, and science-driven decision making. Furthermore, broad and equitable access to diverse artificial intelligence (AI)-ready data repositories is crucial to realize the full potential of AI in responsibly advancing solutions to important scientific and societal problems at national and global scale. In response to these immediate needs, the National Data Platform (NDP) serves as a federated and extensible data ecosystem that fosters collaboration, innovation, and equitable data use, leveraging existing national Cyberinfrastructure capabilities. Connecting distributed computing and data CI systems to provide open access to data from disparate, often siloed repositories and other data sources necessitates a standardized process and services for ingestion, indexing, curation and data analysis. Through a ?removing-the-barriers?' approach to democratizing data, NDP combines needs assessment, co-design, and diversity-aware capacity building with ready-for-scale data CI capabilities, offering data and knowledge management services across the national CI ecosystem. As a national hub of interconnected data hubs, NDP facilitates data discovery and usage, drives responsible AI research and development, fosters scientific understanding, and supports decision-making, policy formation and societal impact. NDP focuses on several case studies in climate and related research areas to evolve data-centric workflows, including wildland fire, and natural and real-time hazards detection and decision making This effort is also supported by National Discovery Cloud for Climate (NDC-C) resources.

Nebula graph operator - will be used for traceroute tool

NEC validator for QA/QC of electrical designs against national electric codes

Building large-scale models for neural decoding and alignment.

Publications: Mehdi Azabou, Vinam Arora, Venkataramana Ganesh, Ximeng Mao, Santosh Nachimuthu, Michael J Mendelson, Blake Richards, Matthew G Perich, Guillaume Lajoie, Eva L Dyer: A unified, scalable framework for neural population decoding, NeurIPS 2023

NetBox is an open source web application designed to help manage and document computer networks. Initially conceived by the network engineering team at DigitalOcean, NetBox was developed specifically to address the needs of network and infrastructure engineers. It encompasses the following aspects of network management: IP address management (IPAM) - IP networks and addresses, VRFs, and VLANs Equipment racks - Organized by group and site Devices - Types of devices and where they are installed Connections - Network, console, and power connections among devices Virtualization - Virtual machines and clusters Data circuits - Long-haul communications circuits and providers Secrets - Encrypted storage of sensitive credentials

Netbox 3

Publications: Netbox 3

netbox-4 helm chart deployment with custom docker image

Publications: netbox-4

netbox-agent is a patched fork of https://github.com/Solvik/netbox-agent that runs on a custom k8s image inside the Nautilus cluster

We work at the intersection of Neuroscience, AI and Large-Scale Data Analysis. We build different kinds of computational models (descriptive, predictive, normative) to help explain the ‘what’, ‘how’ and ‘why’ of information processing in the brain, across domains such as vision, audition, language and multimodal perception.

Publications: https://scholar.google.com/citations?user=ltqwAXYAAAAJ&hl=en

a namespace for prototyping full-stack apps for the Seam Project for UCSC's OPSO and Google Summer of Code, mentor: Mohammad Firas Sada

Nextcloud system namespace. Nextcloud provides file storage, sharing, and many other features.

Nebraska Experimental Testbed of Things (NEXTT) is Nebraska's largest outdoor gigabit wireless experimental testbed. NEXTT is built on seven sites and includes rooftop sites, traffic intersection site, indoor sites, and an agricultural site. NEXTT supports research on 6G, ORAN, dynamic spectrum access, agricultural IoT, and cyber-physical networks. NEXTT is operated by the Cyber Physical Networking (CPN) Lab in the School of Computing at the University of Nebraska-Lincoln.

Publications: https://cpn.unl.edu/publications/

For phd thesis using. The goal is to observe the llama distribution server side gpu device behavior.

Deep Learning techniques for predicting human activity levels from raw accelerometer data. In this project, we develop new Deep Learning based techniques for predicting human activity levels (e.g., sitting, standing, and stepping) from raw tri-axial accelerometer data. The data is collected from a large cohort of human subjects who wore accelerometer devices for seven days of free living. Our goal is to develop accurate methods to predict human activity from these accelerometer data and then use them in downstream human activity and metabolic health correlation analysis. The challenges of this project include working with large volumes of training data (~1 TB) and performing extensive model selection such as neural architecture search and hyperparameter tuning. We are actively using CHASE-CI's persistent storage to store our input and intermediate data. We parallelize the model selection using multiple on-demand GPU virtual machines.

Publications: https://adalabucsd.github.io/DeepPostures/pub/

We conduct NLP research in healthcare and social science domains

We conduct NLP research in healthcare. We extract information from text using state-of-the-art AI methods.

Nvidia node feature discovery - sets labels to nodes for hardware

node-problem-detector aims to make various node problems visible to the upstream layers in the cluster management stack. It is a daemon that runs on each node, detects node problems and reports them to apiserver. node-problem-detector can either run as a DaemonSet or run standalone.

This namespace is made for diffusion experiments. It will be used to explore several aspects including increasing adherence of the models to inputs

Low-income communities in the US have food systems saturated with ultra-processed, hyperpalatable foods– industrially produced “fast foods” that are cheap, convenient, and habit-forming. Meanwhile, fresh food is hard to find. The proliferation of these “food swamps” is a national challenge that is driving twin epidemics of obesity and chronic disease, as well as significant environmental harms and profound health inequities. The urgent need to address this national challenge has been recognized by the National Academy of Medicine, the US Food Department of Agriculture, and the National Science Foundation. The Network Of User-engaged Researchers building Interdisciplinary Scientific infrastructures for Healthy food (NOURISH) will develop technical solutions that help people transform food swamps into healthy food systems. Ultraprocessed and hyperpalatable foods comprise about two-thirds of the US food supply, leaving most Americans to encounter them daily. But food swamps leave their residents with few food options other than these unhealthy products. To solve the problem of food swamps, we must equip responsible business entrepreneurs situated within these communities with data and information for developing and marketing healthy, sustainable foods. This includes information on what consumers in their markets want in terms of taste, convenience, and affordability, as well as information on how to source fresh produce affordably and open a small business. Bringing high technology and advanced data to a cell phone application and online dashboard, NOURISH will enable local food entrepreneurs to grow businesses that produce, prepare, and market food that is naturally appealing, without the industrial production processes used to make hyperpalatable foods. There already exists a large, national network of community-based nonprofit food justice groups seeking this transformation. There is also a vibrant community of philanthropists, investors and social enterprise firms who want to invest in, and support, healthy food businesses in under-resourced communities. NOURISH brings these groups together with scientists to innovate technical solutions that work for everybody. The NOURISH system will connect local food entrepreneurs and investors, equipping them with a high-technology system that accelerates their efforts to transform food swamps. Features of the system will include: ● a national food swamp map, ● crowdsourced data on local consumer food preferences and affordable pricing, ● access to supply chains for fresh foods, ● resources for launching a small food business, and ● a social networking feature that builds and connects stakeholders in healthy food nationwide.

NROTC

A namespace for combined experiments across the National Research Platform's Nautilus Cluster and the FABRIC Testbed using ESnet SENSE and Facility Ports

Running LLM for general use of NRP services. Mostly use h2o one.

The namespace to set up a vector DB for nrp-llm project

Namespace for the weekly NRP security report scanning IPMI interfaces as well as Jupyterhub security.

nrp-sense multus net links connecting FPGAs to L2 SENSE services

PNRP U55C Xilinx FPGA dev environment with Vitis and Vivado tools

Publications: https://github.com/fastmachinelearning/nrp_u55c_benchmark

Multi-domain Knowledge Graph Representation Learning for Digital Twin of Design and Manufacturing

Publications: Journal Publication Part 1: Haolin Fan, Xuan Liu, Jerry Ying Hsi Fuh, Wen Feng Lu, Bingbing Li*, “Embodied Intelligence in Manufacturing: Leveraging Large Language Models for Autonomous Industrial Robotics”, Journal of Intelligent Manufacturing, 2025, Vol. 36: 1141-1157. https://doi.org/10.1007/s10845-023-02294-y Journal Publication Part 2: Haolin Fan, Hongji Zhang, Changyu Ma, Tongzi Wu, Jerry Ying Hsi Fuh, Bingbing Li*, “Enhancing Metal Additive Manufacturing Training with the Advanced Vision Language Model: A Pathway to Immersive Augmented Reality Training for Non-Experts”, Journal of Manufacturing Systems, 2024, Vol. 75: 257-269. https://doi.org/10.1016/j.jmsy.2024.06.007 Journal Publication Part 3: Xuan Liu, John Ahmet Erkoyuncu, Jerry Ying Hsi Fuh, Wen Feng Lu, and Bingbing Li*, “Knowledge Extraction for Additive Manufacturing Process via Named Entity Recognition with LLMs”, Robotics and Computer-Integrated Manufacturing, 2025, Vol. 93: 102900. https://doi.org/10.1016/j.rcim.2024.102900 Journal Publication Part 4: Haolin Fan, Chenshu Liu, Shijie Bian, Changyu Ma, Junlin Huang, Xuan Liu, Marshall Doyle, Thomas Lu, Edward Chow, Lianyi Chen, Jerry Yinghsi Fuh, Wen Feng Lu, Bingbing Li*, “New Era Towards Autonomous Additive Manufacturing: A Review of Recent Trends and Future Perspectives”, International Journal of Extreme Manufacturing, 2025, Vol. 7 (3): 032006. https://doi.org/10.1088/2631-7990/ada8e4 Journal Publication Part 5: Haolin Fan, Junlin Huang, Jilong Xu, Yifei Zhou, Jerry Ying Hsi Fuh, Wen Feng Lu, Bingbing Li*, “AutoMEX: Streamlining Material Extrusion with AI Agents Powered by Large Language Models and Knowledge Graphs”, Materials & Design, 2025, Vol. 251: 113644. https://doi.org/10.1016/j.matdes.2025.113644 Journal Publication Part 6: Haolin Fan, Chenshu Liu, Neville Elieh Janvisloo, Shijie Bian, Jerry Ying Hsi Fuh, Wen Feng Lu, Bingbing Li*, “MaViLa: Unlocking New Potentials in Smart Manufacturing through Vision Language Models”, Journal of Manufacturing Systems, 2025, Vol. 80: 258-271. https://doi.org/10.1016/j.jmsy.2025.02.017 Conference Publication Part 1: Haolin Fan, Jerry Ying Hsi Fuh, Wen Feng Lu, A. Senthil Kumar, and Bingbing Li*, “Unleashing the Potential of Large Language Models for Knowledge Graph Construction: A Practical Experiment on Incremental Sheet Forming”, Procedia Computer Science, 2024, Vol. 232, pp. 1269-1278. International Conference on Industry 4.0 and Smart Manufacturing (ISM 2023), Lisbon, Portugal, November 22-24, 2023. https://doi.org/10.1016/j.procs.2024.01.125 Conference Publication Part 2: Hongji Zhang, Yecheng Jiao, Yizhuo Yuan, Yuanchen Li, Yiqin Wang, Wen Feng Lu, Jerry Ying Hsi Fuh and Bingbing Li*, “Object Detection and Text Recognition for Immersive Augmented Reality Training in Laser Powder Bed Fusion”, Procedia Computer Science, 2024, Vol. 232, pp. 913-923. International Conference on Industry 4.0 and Smart Manufacturing (ISM 2023), Lisbon, Portugal, November 22-24, 2023. https://doi.org/10.1016/j.procs.2024.01.091

This NRP Nautilus JupyterHub server will be an attachment to SuAVE, where users can invoke Jupyter notebooks for additional processing of surveys and image collections. Such notebooks may implement statistical analyses, image processing, machine learning, data mining, semantic image tagging, and other operations.

Network Service Interface node namespace for Automated GOLE Nautilus/UCSD deployment with NSI Safnari + PCE, DDS, OpenNSA and Envoy proxy

Welcome to Enterprise Nucleus Server compose and configuration files!

Open Air Interface noVNC remote development desktop for 5G

oauth2-proxy for securing service behind a haproxy foo.

System namespace - for observablehq - setup 08/28/24

Study of the dynamics of the most eccentric comets in the Oort Cloud

Oklahoma State University IT Knowledge Base Analysis.

Operator Lifecycle Manager manages operators in the cluster

Nvidia Omniverse Digital Twin Experiments, Helm Chart Experiments

The Open Force Field Consortium is composed of academic investigators from the Open Force Field Initiative and sponsoring Industry Partners collaborating to advance open force field science, toolkits, and standards for biomolecular drug discovery.

Publications: Full publications: Development and Benchmarking of Open Force Field v1.0.0—the Parsley Small-Molecule Force Field: https://dx.doi.org/10.1021/acs.jctc.1c00571 Preprints: End-to-End differentiable construction of molecular mechanics force fields: https://arxiv.org/abs/2010.01196 Trained models (Force Fields): OpenFF-2.0.0 "Sage": https://doi.org/10.5281/zenodo.5214478 OpenFF-1.3.1 "Parsley Update": https://doi.org/10.5281/zenodo.5009058 OpenFF-1.3.0 "Parsley Update": https://doi.org/10.5281/zenodo.4118484 OpenFF-1.2.1 "Parsley Update": https://doi.org/10.5281/zenodo.4021623 OpenFF-1.2.0 "Parsley Update": https://doi.org/10.5281/zenodo.3872244

OpenNSA is an implementation of the Network Service Interface (NSI). NSI (Network Service Interface) is a technology agnostic protocol for provisioning network circuits. For more information on NSI, see project page at OGF: https://redmine.ogf.org/projects/nsi-wg OpenNSA is currently in a state of heavy development, and many features are only partially implemented.

This is a namespace for SENSE development of new opennsa versions and applications

Opertators deployed by OLM from operator hub in automated way

Caching technology deployed on Nautilus cluster: Stashcache is a caching infrastructure based on the XrootD software. We deploy stashcache containers at grid sites and also in the Internet backbone. The objective is to reduce latency for scientific datasets (open and private) that are accessed at several computing sites. At a computing site the "nearest" cache based on GeoIP is picked and accessed. LIGO and the caching technology: The LIGO experiment has computing resources located at several location in the US and above. Moreover it can also access the VIRGO computing resources located in Europe. LIGO uses OSG powered technology glideinWMS to run workflows on its own computing resources, VIRGO resources and opportunistic resources. Given the distributed nature of its computing it needs to be able to securely access (Only members of the LIGO collaboration can access these datasets) its input data. The secure caching infrastructure deployed all over using kubernetes provides this. Image credit: LIGO/T. Pyle

Namespace for deploying glideinwms-frontends as a service to multiple scientific communities. An example of this is the UCI FE used by ATLAS

OSG Grid Infrastructure Laboratory (GIL) is dedicated to explore new distributed system technologies.

Opportunistic use of resources by the IceCube collaboration using HTCondor pilots through the k8s provisioner.

Publications: Auto-scaling HTCondor pools using Kubernetes compute resources, https://arxiv.org/abs/2205.01004 The anachronism of whole-GPU accounting, https://arxiv.org/abs/2205.09232 https://icecube.wisc.edu/science/publications/

Opportunistic use of PNRP resources by the LIGO/IGWN collaboration using the HTCondor provisioner.

Publications: Will update later

Jobs running under the HTCondor-provisioned resources through the OSPool. More details in the OSPool accounting system.

Opportunistic use of resources by the PATh OSPoool using HTCondor pilots through the k8s provisioner.

Namespace to hold OSDF services deployed using Nautilus

Publications: StashCache: A Distributed Caching Federation for the Open Science Grid

Experimentation associated with off-grid, decentralized LoRa-based communications mesh

The JusticeTech initiative at OSU aims to analyse court documents related to eviction and truancy as well as create software to support different parties.

Jupyter Lab pilot for researchers and students associated with the University of Oklahoma This namespace is supporting research and classroom teaching as a pilot through the University of Oklahoma Libraries.

Publications: Presentation at Coalition of Networked Information 2023, Educause podcast interview 2023 Presentation at University of Oklahoma Academic Tech Expo 2022, Presentation at GPN 2022 meeting, OU Libraries 2022 UL Week Lightning Talk

Pilot workspace for University of Oklahoma research groups batch connect with GPUs as well as CPUs for AI/ML.

Publications: Presentation at University of Oklahoma Academic Tech Expo 2022, Presentation at GPN 2022 meeting

Various testing efforts and socialization of the platform Utilizing namespace for undergraduate CS student workers and Graduate Assistants to gain experience with Kubernetes

Publications: Presentation at University of Oklahoma Academic Tech Expo 2022, Presentation at GPN 2022 meeting, OU Libraries 2022 UL Week Lightning Talk

NRP onboarding for AI and containerization with Digital Scholarship and Data Services Graduate Assistant within University of Oklahoma Libraries.

Omniverse Farm Queue and Omniverse Farm Agent allow you to run tasks in the background, and to run automated jobs defined by you or others. They can be used for a number of different use cases, including: Rendering frames or movie clips Sharing resources across multiple machines Automating repetitive, or time-consuming tasks, such as batch file conversion or validating asset naming conventions Generating turntable-style asset previews Generating levels of details for assets Simulating physics, or baking fluid caches Generating USD scenes to train machine learning models Exporting BIM data from AEC projects as USD layers etc. Both Omniverse Farm Queue and Omniverse Farm Agent are designed from the ground up to be infrastructure-agnostic, and embrace the microservice architecture for flexibility and scalability. This means they designed to run on typical workstations, bare-metal servers or even advanced Cloud platforms such as Kubernetes.

Overleaf LaTeX editor - A web-based collaborative LaTeX editor

A namespace for all Intel Tofino P4 related materials, including pipelines and training materials.

Examining Longitudinal Changes in Accelerometer-Measured Physical Activity in Preventing Cardiovascular Disease with Novel Function Data Analysis Approaches

pace-ucicl

panostream 360 viewer

Study of primordial black hole capture in the solar system

Perfsonar deployment - active network measurements

Development of causal inference, survival analysis methodology

Distributed machine learning: We develop distributed machine learning setting like federated learning in order to optimize communication and computation overhead of the training.

Publications: https://scholar.google.com/citations?user=mjfYEY8AAAAJ&hl=en

Physics guided machine learning for prescribed burn simulation

Phillp's name space for testing SNN for Loihi 2 and using tutorials, using mainly PyTorch, CUDA.

The Piraeus Operator manages LINSTOR clusters in Kubernetes.

The Postgres Operator delivers an easy to run highly-available PostgreSQL clusters on Kubernetes (K8s) powered by Patroni. It is configured only through Postgres manifests (CRDs) to ease integration into automated CI/CD pipelines with no access to Kubernetes API directly, promoting infrastructure as code vs manual operations.

a namespace for prototyping full-stack apps for the Seam Project for UCSC's OPSO and Google Summer of Code, mentor: Mohammad Firas Sada

UCSC CSE Capstone: Deep Learning for NLP and Multi-Modal Applications

This is the heterogeneous memory project group supported by UCSD and PRISM center. We are PhD students from multiple groups led by PI Tajana Rosing, Jishen Zhao, Dean Tullsen, and Steve Swanson. We aim to understand characteristics and optimization techniques with all kinds of memory and system technologies, such as CCIX and CXL.

Publications: https://arxiv.org/pdf/2411.02814

PRISM EGL Desktop ......................................

caching experiments using AMD/Xilinx Alveo U55C FPGAs

prism-jupyterlab notebook server ...................

Unifying Program Representation at Source and Assembly Code Levels for IC Design Automation

this is a pilot project for testing NRP will start with containers that have worked on sol supercomputer at Arizona State University

3D Modeling at Wadi el-Hudi In 2019, CSUSB faculty, staff, and students undertook a comprehensive photogrammetric survey of archaeological sites in a region of Egypt’s Eastern Desert at Wadi el-Hudi as part of the Wadi el-Hudi Expedition (www.wadielhudi.com). In this survey they took more than 14600 photographs, which constitute the data for making 3D meshes using Agisoft Photoscan. Because of the size of data for the models, we are using PRP Chase-CI cluster computing to render the 3D models.

Publications: "None"

pyCoral example

PypeIt is a Python package for semi-automated reduction of astronomical, spectroscopic data. Its algorithms build on decades-long development of previous data reduction pipelines by the developers (Bernstein, Burles, & Prochaska, 2015; Bochanski et al., 2009). The reduction procedure -- including a complete list of the input parameters and available functionality -- is provided as online documentation hosted by Read the Docs, which is regularly updated. (https://pypeit.readthedocs.io/en/latest/). Release v1.0.3 serves the following spectrographs: Gemini/GNIRS, Gemini/GMOS, Gemini/FLAMINGOS 2, Lick/Kast, Magellan/MagE, Magellan/Fire, MDM/OSMOS, Keck/DEIMOS (600ZD, 830G, 1200G), Keck/LRIS, Keck/MOSFIRE (J and Y gratings tested), Keck/NIRES, Keck/NIRSPEC (low-dispersion), LBT/Luci-I, Luci-II, LBT/MODS (beta), NOT/ALFOSC (grism4), VLT/X-Shooter (VIS, NIR), VLT/FORS2 (300I, 300V), WHT/ISIS.

Publications: Prochaska J., Hennawi J., Westfall K., Cooke R., Wang F., Hsyu T., Davies F., et al., 2020, JOSS, 5, 2308. doi:10.21105/joss.02308

qaic-jump for Qualcomm AI 100 Ultra cards in NRP The installation is fully self contained but requires additional environment variables to locate the tools.

We are building a Quantum Data Hub System to amplify the value of foundry data through data science. Highlights of the System: Experiment design process capture and management, Data modeling, storage, and access Collaborative analysis interfaces.

Publications: "None"

Designing scalable passive reflective metasurfaces for wireless networking and sensing

This project is for AI model design for channel prediction.

A namespace for the development of Qualcomm Cloud AI 100 using the Cloud AI 100 SDK

Intel® Quartus® Prime design software development environment

QUICK (Quantum Interaction Computational Kernel) is a GPU-enabled open-soruce quantum chemistry software supported by NSF through a CSSI Elements and Frameworks award. It is the default compute engine for quantum mechanical (QM) and mixed quantum/classical (QM/MM) molecular dynamics (MD) simulations in the widely used AMBER biomolecular simulations package. We are exploring mixed FP64, FP32, and FP16 algorithms on a variety of GPU platforms to accelerate physics based QM/MM MD simulations and to provide distributed high-throughput capabilities for rapid training data generation of QM based machine learning (ML) models for molecular simulations.

Research for solving problems in theory of numbers, e. g., lattice reduction, shortest vector problem, prime factorisation etc.

Publications: https://tashfeen.org

This workspace is used for research development in AI and Machine Learning for students at Responsible and Reliable AI Lab.

Publications: https://scholar.google.com/citations?hl=en&user=9rpkTSkAAAAJ&view_op=list_works&sortby=pubdate

Racelab focuses on distributed system, cloud computing and IoT research.

Ray cluster operator

We're interested in Machine Learning & AI and their applications to fundamental problems in medicine and biology. Our interests are in building state-of-the-art models, making theoretical advances, as well as making fundamental discoveries in biology and medicine. We are currently working on image reconstruction for medical scans, medical image generation, disease progression modelling, as well as building MRI/PET simulators.

Publications: See here: https://razvanmarinescu.com/

My research interest is human-centered machine learning. My goal is to build responsible machine learning tools with humans in the loop, including developing 1) robust training methods to deal with noisy human-generated inputs, 2) fair and accountable machine learning treatments to better serve our society, and 3) incentive-compatible data collection mechanisms. The central question associated with my work is learning from dynamic and noisy data. I am fascinated by the question of how much we are able to learn in a weakly supervised setting (e.g., with noisy supervision, low-quality/biased training inputs, strategic manipulations, etc).

Publications: Please see: http://www.yliuu.com/research/index.html

Daily email to get cpu/gpu hours used in past 24 hours per namespace, as well as security report on IPMI and Jupyterhub vulnerabilities

Current: This server is for the research group consisting of Dr. Edray Goins, professor of mathematics at Pomona College, Dr. Youngsu Kim, assistant professor of mathematics CSU San Bernardino, Tesfa Asmara, undergraduate student at Pomona College working on a project in game theory. Past: JupyterHub for 2023 Pomona Research in Mathematics Experience, https://pages.pomona.edu/~ehga2017/prime.html This is a seasonal project and will end by the end of Aug 2023. Past: JupyterHub for 2022 Pomona Research in Mathematics Experience, https://pages.pomona.edu/~ehga2017/prime.html This is a seasonal project and will end by Aug 2022.

USRA Research Institute for Advanced Computer Science (RIACS) Labs project hosts interactive educational materials in advanced data science and quantum computing.

Publications: N/A

For testing GPUs, and learning how to use PyTorch to use it.

exploring RL on motion imitation, for leg locomotion control. We will explore complex legged locomotion for A1 robot and try to learn them together, and apply on complex terrains.

Publications: Ruihan Yang*, Minghao Zhang*, Nicklas Hansen, Huazhe Xu, Xiaolong Wang. Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers. International Conference on Learning Representations (ICLR), 2022 (Spotlight Presentation).

Multi-task learning is a very challenging problem in reinforcement learning. While training multiple tasks jointly allow the policies to share parameters across different tasks, the optimization problem becomes non-trivial: It is unclear what parameters in the network should be reused across tasks, and the gradients from different tasks may interfere with each other. Thus, instead of naively sharing parameters across tasks, we introduce an explicit modularization technique on policy representation to alleviate this optimization issue. Given a base policy network, we design a routing network which estimates different routing strategies to reconfigure the base network for each task. Instead of creating a concrete route for each task, our task-specific policy is represented by a soft combination of all possible routes. We name this approach soft modularization. We experiment with multiple robotics manipulation tasks in simulation and show our method improves sample efficiency and performance over baselines by a large margin.

Publications: Ruihan Yang, Huazhe Xu, Yi Wu, Xiaolong Wang. Multi-Task Reinforcement Learning with Soft Modularization. Conference on Neural Information Processing Systems (NeurIPS), 2020. Nicklas Hansen, Xiaolong Wang. Generalization in Reinforcement Learning by Soft Data Augmentation. International Conference on Robotics and Automation (ICRA), 2021. Qiang Zhang, Tete Xiao, Alexei A. Efros, Lerrel Pinto, Xiaolong Wang. Learning Cross-domain Correspondence for Control with Dynamics Cycle-consistency. International Conference on Learning Representations (ICLR), 2021 (Oral Presentation). Nicklas Hansen, Rishabh Jangir, Yu Sun, Guillem Alenyà, Pieter Abbeel, Alexei A. Efros, Lerrel Pinto, Xiaolong Wang. Self-Supervised Policy Adaptation during Deployment. International Conference on Learning Representations (ICLR), 2021 (Spotlight Presentation).

We explore RL for dexterous manipulation. We train our model in sim and apply on real.

Publications: Yuzhe Qin*, Yueh-Hua Wu*, Shaowei Liu*, Hanwen Jiang*, Ruihan Yang, Yang Fu, Xiaolong Wang. DexMV: Imitation Learning for Dexterous Manipulation from Human Videos. arXiv, 2021. Yuzhe Qin, Hao Su*, Xiaolong Wang*. From One Hand to Multiple Hands: Imitation Learning for Dexterous Manipulation from Single-Camera Teleoperation. arXiv, 2022.

System namespace - rook ceph, the storage system component

US Central ceph pool, System namespace - rook ceph, the storage system component

Ceph eastern zone, System namespace - rook ceph, the storage system component