If you're interested in distributed computing, the APeX Summit (Academy, Parsl, Globus Compute) is approaching. This meeting brings together researchers, developers, and cyberinfrastructure experts worldwide to discuss experiences and define future directions. parsl-project.org/parslfest/ap...

AI agents are showing up everywhere in science right now, but deploying them on real research infrastructure presents unique challenges: e.g., federated authentication, restrictive networks, diverse compute, and systems that aren't always up. Paper: doi.org/10.2139/ssrn...

Lens methods let researchers watch an LLM's predictions form layer by layer, but apply them to larger LLMs has been problematic. Towards this end, the team has developed OmniLens: low-rank translators cut trainable parameters by up to 98%, and Subset-KL training slashes peak memory by up to 70%.

The LLM Hackathon for Applications in Materials and Chemistry is now the Open Scientific Intelligence Hackathon! Hosted by @benblaiszik.bsky.social, our next event is Oct 21–22, 2026 — free, hybrid, worldwide, now covering physical sciences & math too. Register: llmhackathon.github.io

Open Scientific Intelligence (OSI) Hackathon for the Physical Sciences & Mathematics

The Open Scientific Intelligence (OSI) Hackathon runs October 21–22, 2026. Formerly the LLM Hackathon for Materials Science & Chemistry — now covering agents, datasets, models, and scientific software...

llmhackathon.github.io

Congratulations to Dr. J. Gregory Pauloski on winning the SIGHPC 2026 Doctoral Dissertation Award for his dissertation, "Programming the Continuum: Towards Better Techniques for Developing Distributed Science Applications"! 🎉 www.sighpc.org/opportunitie...

Meeting Your Needs - 2026 Award Winner

Dr. J. Gregory Pauloski,a graduate of the University of Chicago Recognized for his dissertation "Programming the Continuum: Towards Better Techniques for Developing Distributed Science Applications” ...

sighpc.org

Our DSI summer student, Ben Mechels, recently presented his research on “Transit-Layer Tensor Representation for Serverless Inference” at Argonne National Labs! www.anl.gov/event/transi...

Congratulations to our postdoc Suman Raj on being accepted to the Early Careers Program at SC '26! sc26.supercomputing.org/program/earl...

Early Career • SC26

The Early Career program is a series of workshop-style sessions and pre-conference webinars for people in the early stages of their careers.

sc26.supercomputing.org

Congrats to @kylechard.bsky.social and @benblaiszik.bsky.social on being a part of the new NSF Materials Innovation Platforms at @uwmadison.bsky.social! The initiative will advance materials research through AI, automation, and expand access to cutting-edge facilities. www.nsf.gov/news/nsf-sta...

NSF stands up 2 Materials Innovation Platforms with $50M investment

The U.S. National Science Foundation is investing $50 million in research infrastructure to enable the discovery and development of materials that can withstand extreme conditions, from lightweight…

nsf.gov

We’re thrilled to be part of PoLARIS, a new AI- and robotics-powered laboratory supported by a $20M NSF award! The project will expand access to advanced materials research for scientists nationwide. cs.uchicago.edu/news/remotel...

Remotely Operated, Robotic Lab Receives $20 Million National Science Foundation Grant - Department of Computer Science

Tomorrow’s clean energy, advanced manufacturing, healthcare, and national defense rely on developing novel, next-generation soft materials – things like liquids, polymers, gels, colloids, foams, and g...

cs.uchicago.edu

We're excited to have our summer students Alex Zhang and Eva Neves from UChicago Lab Schools, Francesca Kang from Kenwood Academy, Noah Runesha from Princeton, and Ben Mechels from UMN join Globus Labs. Wishing them a summer of curiosity, collaboration, and discovery!🎉

We're heading to #ISC26 in Hamburg, June 22–26! 🚀 Join us for a hands-on Academy tutorial, Kyle's talk on agentic AI, and our Icicle talk on scalable metadata indexing for HPC file systems. Stop by the Globus booth F11 and see our ISC schedule: www.globus.org/events/isc-h...

Thrilled to share that last week, Owen Price Skelly presented at MLCM-26 on how Garden and RootStock help materials scientists use state-of-the-art MLIPs on HPC. Check them out: thegardens.ai garden-ai.github.io/rootstock/

SciFM’26 at UChicago was a blast! 🎉 Proud to see our PhD student Song Oh present “Searching the Pareto Front,” exploring multi-objective retrieval for heterogeneous scientific modalities and late-fusion embeddings for scientific discovery. www.scifmconferences.org

SciFM Conferences | Scientific Foundation Models and AI Agents

scifmconferences.org is the official website of the Scientific Foundation Models Conferences (SciFM24, SciFM25, & SciFM26) organized by the University of Chicago, the Michigan Institute for Computatio...

scifmconferences.org

We will be giving a tutorial on Academy at ISC on June 22nd!

Kyle, Ian, Yadu, and Alok will give a tutorial on Academy at IPDPS on May 26th

We're excited to share that Gus Ellerm has joined Globus Labs as a postdoc 🎉 He earned his PhD from the University of Canterbury working on provenance in scientific systems and developing self-updating research publications integrated with LLMs.

Introducing Groundhog 🦫☀️, a package to make working with HPC Python functions across servers delightful - especially if you need multiple tries to get it right! groundhog-hpc.readthedocs.io/en/latest/

Globus Streaming in now available at https://www.globus.org/streaming-data. Go check it out!

Ian gave two talks at SCA/HPCAsia 2026: "System Requirements for Scalable Agentic AI" and "Agentic Systems for Vaccine Design" Great discussions on how agentic systems can transform scientific computing and accelerate discovery. www.sca-hpcasia2026.jp

Our paper "Empowering Scientific Workflows with Federated Agents" has been accepted to IPDPS'26! We introduce Academy - middleware for deploying autonomous agents across HPC systems, experimental facilities & data repositories. github.com/academy-agen...

GitHub - academy-agents/academy: Build and deploy stateful agents across federated resources

Build and deploy stateful agents across federated resources - academy-agents/academy

github.com

New work! We show that machine-learned operators (MLOs) for PDEs struggle with accurate zero-shot multi-resolution inference. Popular fixes don’t solve it, but multi-resolution training offers robust, cost-effective modeling across scales! https://github.com/msakarvadia/operator_aliasing

Thank you to Samuel Kounev and Nikolas Herbst for hosting Ian and Kyle at the SOS project meeting (dfg-sos.de) and the University of Würzburg.

Alok gave a talk on "Building Scalable Agentic Systems for Science with Academy" at the Agentic AI for User Facilities workshop (indico.physics.lbl.gov/event/3209/)

Agentic AI for User Facilities

Agentic AI, systems that autonomously sense, decide, and act, is rapidly transforming operations across the Department of Energy’s user facilities. This two-day workshop will convene experts from ligh...

indico.physics.lbl.gov

Ian gave an invited talk at the University of Würzburg on "Accelerating Discovery with Intelligent Agents"

We're excited to share that Suman Raj has joined Globus Labs as a postdoc.

LSHBloom, a memory-efficient way to deduplicate documents at scale - critical for LLM training pipelines - has been included in the Datasketch package for easy use. See the results of our Exascale document deduplication and more details about LSHBloom in the paper at: arxiv.org/abs/2411.04257

@danielgrzenda.bsky.social presented his work on MOFA at Climate Frontiers (climate.uchicago.edu/climate-fron...)

Hai Duc Nguyen presented a poster to the Argonne's Postdoctoral Research and Career Symposium 2025 on "Resilient Execution of Distributed X-Ray Image Analysis Workflows" on Friday, 10/31 tinyurl.com/4av9xnsx

Frontiers | Resilient execution of distributed X-ray image analysis workflows

Long-running scientific workflows, such as tomographic data analysis pipelines, are prone to a variety of failures, including hardware and network disruption...

tinyurl.com

Hai Duc Nguyen’s paper, “Efficient Performance Guarantees for Function-as-a-Service with Cloud Allocators,” has been accepted for publication at Middleware 2025.

Check out the new Chemical Foundation Model paper done in collaboration with UofM: arxiv.org/pdf/2510.18900