Globus Labs

Maxime Gonthier presented his work on "Failure-Aware and Deadline-Constrained Scheduling for Mixed-Criticality HPC Workloads" (inria.hal.science/hal-05190988...) to the Data & AI Systems Research Section at ORNL.

Our work in collaboration with the Gagliardi group: “Cartesian Equivariant Representations for Learning and Understanding Molecular Orbitals” accepted to the Proceedings of the National Academy of Sciences (PNAS)
chemrxiv.org/engage/api-g...

@danielgrzenda.bsky.social @nathaniel-hudson.bsky.social

conference.eresearch.edu.au/enabling-res...

conference.eresearch.edu.au/solving-data...

Kyle is presenting in the "Solving Data Management Challenges with Globus – from Zero to Hero in one day" tutorial and participating in the Enabling Research Connectivity: The Strategic Role of Data Mobility in an Integrated National Research Infrastructure Ecosystem. BoF at eResearch Australasia.

@ianfoster42.bsky.social gave a talk “Scaling Reasoning, Scaling Science: Engineering an AI-Native Scientific Discovery Platform” to Google Cloud, presenting a vision for how frontier-scale reasoning models can be integrated with scientific infrastructure to accelerate discovery.

Lasting Innovation in Data Movement: SC25 Test of Time Award • SC25 The annual Test of Time Award recognizes an outstanding paper that has deeply influenced the HPC discipline.

🎉The SC25 Test of Time Award goes to "The Globus Striped GridFTP Framework and Server"! Their pioneering work revolutionized secure, high-performance data movement, shaping scientific discovery for decades. Congrats to the team! #GridFTP #Supercomputing sc25.supercomputing.org/2025/10/last...

Steering an Active Learning Workflow Towards Novel Materials Discovery via Queue Prioritization Generative AI poses both opportunities and risks for solving inverse design problems in the sciences. Generative tools provide the ability to expand and refine a search space autonomously, but do so a...

Check out the paper: arxiv.org/abs/2509.25538

🧪🚀 Exciting breakthrough! Our new paper presents a an algorithm that combines #ML & #ActiveLearning to turbocharge discovery of novel materials. By smarter candidate selection, our method doubles high-performing discoveries for carbon capture molecules! #MaterialsScience

iCNS Distinguished Lecture: Scaling Autonomy: Federated Agentic Middleware and Serverless Systems for Scientific Computing Presented by Kyle Chard, Research Associate Professor, University of Chicago. Hosted by the Department of Electrical and Computer Engineering.

👨‍💻 Don’t miss the iCNS Distinguished Lecture by Dr.
@kylechard.bsky.social! Learn how federated agentic middleware & serverless systems are revolutionizing scientific computing—integrating diverse resources, and enabling high-throughput discovery.

www.stevens.edu/events/icns-...

Haibin Lai and Maxime Gonthier presenting at ICPP!

Congratulations to Globus Labs alum @tkurihana.bsky.social on his new position as a Senior Researcher at Fujitsu Research Converging Technologies Laboratory! 🎉🎊

Experiences with Model Context Protocol Servers for Science and High Performance Computing Large language model (LLM)-powered agents are increasingly used to plan and execute scientific workflows, yet most research cyberinfrastructure (CI) exposes heterogeneous APIs and implements security ...

Check out our latest preprint (arxiv.org/abs/2508.18489), we explore using the Model Context Protocol (MCP) to connect AI agents with HPC and scientific services, making research cyberinfrastructure (CI) discoverable, invokable, and composable.

ParslFest 2025 was a huge success! Thank you to all 40 speakers and all who attended in person and online: parsl-project.org/parslfest/pa...

Our paper “Steering an Active Learning Workflow Towards Novel Materials Discovery via Queue Prioritization” by Marcus Schwarting, Logan Ward, Nathaniel Hudson, and others is nominated for Best Paper at this year’s e-Science conference! 🎉

🚀 Join us for ParslFest 2025!
Meet researchers, developers, and cyberinfrastructure experts to share experiences and explore the latest in Parsl and Globus Compute.

📅 Aug 28–29 | 🏙️ Chicago & 🌐 Virtual (free)
🔗 Register now: parsl-project.org/parslfest/pa...

That's a wrap on GLabs Summer '25! 🎓🚀 Huge thanks to our summer students for your energy at GLabs: Daniel Babnigg, Edwin Ma, Jin Das, Naomi Kolodisner, Santhusha Mudannayaka, Shivalika Chaudhary, Suryatejas Appana, and Rui Wang. Can't wait to see what you build next!

The final version is now available in the Journal of Parallel and Distributed Computing: www.sciencedirect.com/science/arti...

Congrats to postdoc Maxime Gonthier!🎉 His work “A scheduler to foster data locality for GPU and out-of-core task-based linear algebra applications” has been accepted for publication at the Journal of Parallel and Distributed Computing.

Jin Das, a student from Columbia with the DSI Summer Lab Program, has been working with @danielgrzenda.bsky.social. She’ll present “TideCast: Benchmarking Deep Learning Models for Water Level Prediction” at the DSI Research Symposium on Friday. It's been great working with you this summer, Jin!

Naomi Kolodisner, a University of Arizona student, has been working with Alok/Greg on genomics workflows using Academy and Parsl. She'll present her work: "Adaptive Tool Selection in a Scalable Genomics Pipeline" tomorrow at the CI-Compass Fellowship Summer Symposium. Awesome job this summer Naomi!

@ianfoster42.bsky.social will be speaking on the AI & Science Panel (kdd2025.kdd.org/kdd-ai-scien...) this week at #KDD25. Don't miss it! 😉

The Trillion Parameter Consortium Convening AI leaders globally from industry, academia, and national laboratories

🚀 Last week at #TPC25 (tpc25.org), @ianfoster42.bsky.social delivered two keynotes and also joined the closing panel:

1️⃣ Scaling Reasoning, Scaling Science: Engineering an AI-Native Scientific Discovery Platform
2️⃣ A Case Study of the System Software/Middleware Needs for Agents

Excited to see our "MProt-DPO framework for protein design featured on SciShow" as a highlight Aurora application!
🎥 Watch the segment: www.youtube.com/watch?v=pC3Q...

Thrilled to present "Globus as a Case Study in Sustainable Research Software Development" at the NSF CSSI/CyberTraining/SCIPE PI meeting plenary session!
Check it out to see how long-term vision and community collaboration drive sustainable research infrastructure:
confmeet.github.io/2025NSFCyber...

Kyle Chard presented two cutting-edge projects pushing the boundaries of intelligent infrastructure and AI for science at #TPC25! 🎉

🔹 Academy: Empowering scientific workflows with federated agents
🔹 Diamond: Democratizing large foundation model training for science

Congratulations Dr. Baughman on successfully defending your dissertation! Matt has been a such an important and valued member of our lab these past few years and has done amazing work while here. We're so proud of what you've accomplished and are excited to see what you do next! 🎉🎓

Will It Blend? Mixing Numerical and Machine-Learned Physics Quantities for Accurate on-the-Fly Surrogate Modeling Learning and deploying inexpensive replacements for computationally expensive subroutines “on-the-fly” (OTF) during a dynamic simulation offers potential advantages and unique drawbacks. I...

Mike Tynes paper "Will It Blend? Mixing Numerical and Machine-Learned Physics Quantities for Accurate on-the-Fly Surrogate Modeling" was published in ICCS 25 (link.springer.com/chapter/10.1...) The paper introduces Proxima+Blend to reduce artifacts when using ML surrogates. Great work Mike!

We're wishing Matt Baughman the best of luck as he gets ready to defend his dissertation this Friday. He will be giving a presentation titled: "Corralling the Computing Continuum: Enabling Multi-System Workflows with Serverless Computing." Good luck Matt!

🚀 On July 24, our very own Matt Baughman is speaking at Princeton Plasma Physics Lab!

His seminar for the Computational Sciences Department, “Corralling the Computing Continuum: Enabling Multi-Site Workflows with Serverless Computing,” dives into the future of distributed workflows.