Lmst

For a more general look at foundation models and their application in science. Check out this brand new podcast interview with Ian Foster of Argonne National Laboratory and the University of Chicago. Full episode available here: https://scienceinparallel.org/2025/02/ian-foster-exploring-and-evaluating-foundation-models/ or on your favorite 🎙️ platform. #HPC #AI

Foundation models, an AI type that includes LLMs, could help researchers fortify defenses against infectious diseases. Arvind Ramanathan of Argonne National Laboratory and his team are applying their award-winning strategy for modeling SARS-CoV-2 evolution toward a more general approach for understanding viruses and bacteria. https://ascr-discovery.org/2025/02/disease-watch/ #HPC #LLM #infectious_disease

Blobs of color: blue, purple and green. Caption/credit: A detail from a genome-scale language model, or GenSLM. Researchers can use a GenSLM to track the evolution of COVID-19 and other infectious agents. Image courtesy of Arvind Ramanathan/Argonne National Laboratory.

Superdiamonds could represent a novel, harder form of carbon that might exist in the cores of exoplanets. Ivan Oleynik and his University of South Florida colleagues have used Oak Ridge National Laboratory's Frontier supercomputer to study how this structure could form. The work is supporting experiments at Lawrence Livermore National Laboratory's National Ignition Facility. https://ascr-discovery.org/2025/02/hard-target-2/ #HPC

Black background with green and blue blobs rimmed with yellow

A rendering of a million-atom molecular dynamics simulations of nucleation in hot dense carbon liquid. Image courtesy of Ivan Oleynick, University of South Florida.

Fourth-year #DOECSGF recipient Rahul Sahay of Harvard University is using #quantum behavior to understand matter's phases and develop strategies to create new ones in the laboratory. https://deixismagazine.org/2025/02/sculpting-nature/

Large circle which includes a series of smaller circles (white, blue, gold and black) linked by lines. Caption: A schematic of a quantum circuit used to study quantum gravity. A one-dimensional magnet enters the circuit at the circle’s circumference and a model of gravity is produced at the black dots. The blue, white, and yellow circles represent the quantum gates making up the circuit. (Image: Rahul Sahay, Mikhail D. Lukin, Jordan Cotler, arXiv:2401.13595.)

Cosmic rays—accelerated atomic nuclei driven by astrophysical phenomena—provide ways to study how galaxies evolved. Using exascale computing power and an #INCITE award researchers are modeling both large- and small-scale behavior in these systems and developing a new model that accounts for unexplained turbulence. https://ascr-discovery.org/2025/02/untangling-the-cosmos/ #HPC #OLCF #ALCF

Brightly-colored lines with sharp kinks. Colors include golds, magenta, blues and teal.
A zoom-in panel demonstrating the small scale complexity in high-resolution simulations of a slice of 3D simulations of cosmic rays. Colors show the electric current’s magnitude. Image courtesy of Drummond Fielding/Cornell University.

Simulations to study new ways to catalyze hydrogen production could help with energy storage. Harvard University's Boris Kozinsky and his team are using machine learning to accelerate these complex calculations. https://ascr-discovery.org/2025/01/carbon-free-hydrogen/ Their #INCITE award includes allocations at both #ALCF and #OLCF. #HPC

Field of gray and white spheres with a couple red ones

An extreme-scale reactive molecular dynamics simulation. Image courtesy of Boris Kozinsky, Harvard University.

New 🎙️ episode: the Science in Parallel podcast is wrapping up Season 5 with computational scientists Mansi Sakarvadia of the University of Chicago and Josh Vermaas of the MSU-DOE Plant Research Laboratory discussing how AI and the recent Nobel Prizes are shaping science. https://scienceinparallel.org/2025/01/computational-scientists-discuss-the-2024-nobel-prizes/ #DOECSGF #HPC

Alex Ballow of Montana State University-Bozeman uses category theory, a form of universal mathematical language, to probe patterns in physics that could advance #quantum computing, condensed matter physics and more.
https://deixismagazine.org/2025/01/categorical-imperative/# The #DOECSGF recipient completed practicum research at Oak Ridge National Laboratory.

Gold device with many cables and cylinders

Caption and credit:
The Quantum Systems Accelerator at Berkeley Lab. DOE CSGF recipient Alex Ballow studies math that informs what happens in quantum systems. (Photo: Thor Swift/Berkeley Lab.)

New on the Science in Parallel podcast 🎙️ : @anil talks about AI's Nobel Prize year and his book Why Machines Learn. Listen on your favorite platform or here: https://scienceinparallel.org/2024/12/anil-ananthaswamy-ais-nobel-moment/ #HPC #AI #nobel2024

The human brain's trillions of connections remain largely uncharted. But a team from Argonne National Laboratory and Harvard University is using advanced electron microscopy and #ALCF's hardware and machine learning tools to map neurons' links to better understand how they interact https://ascr-discovery.org/2024/12/connecting-the-neurodots/ #HPC

Brightly colored oblong shapes linked by a densely packed fibrous network.

Caption: A selected subset of neurons from electron-microscopic brain-tissue images reconstructed on Aurora. The picture was generated using a technique akin to AI imaging, called a convolutional neural network. Image courtesy of Argonne National Laboratory from original data provided by Lichtman Lab, Harvard University.

An initial interest in cars led
Victor Zendejas Lopez to study turbulence using computational fluid dynamics at Caltech and Lawrence Livermore National Laboratory. Read more about his #DOECSGF-supported
research and how mentoring has shaped his career. https://deixismagazine.org/2024/11/predicting-chaos/ #HPC

As large language models get bigger, they're increasingly difficult to train. But a team led by Abhinav Bhatele of University of Maryland has developed the open-source #AI training framework, AxoNN, which distributes the work across multiple GPUs. https://ascr-discovery.org/2024/11/we-the-ai-trainers/ The work is nominated for a 2024 ACM, Gordon Bell Prize at #SC24 and has been supported by an #INCITE allocation at #OLCF and #ALCF.

An artist’s rendering of an artificial neural network. A large language model, or LLM, is built of multiple nodes arranged in layers, each node programmed to mimic a neuron. A user feeds batches of training data to an input layer, and the LLM processes and advances each batch through any number of hidden layers and produces the results in an output layer. Image courtesy of Akritasa via Wikimedia Commons.

Lithium-ion batteries rely on stable, functional electrolytes to store energy. The University of Michigan's Venkat Viswanathan is leading a team building a foundation model to design better electrolyte recipes that could be useful for many chemistry applications. https://ascr-discovery.org/2024/10/ai-turbocharge/ #HPC #renewableenergy #ALCF #LLM

Flow chart. Left side is large blue field with several organic molecules, middle has three arrows leading to a checkerboard field of outcomes that show either property prediction or molecular generation.

Caption/credit: Foundation models for molecule design for energy storage applications. Image courtesy of Anoushka Bhutani, University of Michigan.

During the pandemic, Margaret Cheung pivoted to a career modeling the molecular mechanisms of disease transmission at Pacific Northwest National Laboratory. Using a digital twin of the process, she's helping to grapple with complex biological data and find new strategies to disrupt infections. https://ascr-discovery.org/2024/10/pandemic-preparedness/ #HPC #OLCF #ALCF #NERSC

Nearly spherical neon-green cells on gray background. Caption: A colorized scanning electronic microscopic image of Prochlorococcus, a beneficial organism whose functions can be compromised by viral infection. Image courtesy of Chisholm Lab/MIT via Creative Commons.

New 🎙️ episode: Paulina Rodriguez of
George Washington University discusses her quest to become an authentic scientist and her applied math career focused on the credibility of computational models. Check out the whole episode: https://scienceinparallel.org/2024/10/paulina-rodriguez-building-credibility-and-authenticity/ #HPC

Luis Rangel DaCosta of UC Berkeley has used machine learning to speed up the tedious analysis of high-resolution, atomic-level images. Read more about Construction Zone, the software package he and his colleagues developed "like computer-aided design, but for atoms." https://deixismagazine.org/2024/09/exploring-electrons/ #HPC #DOECSGF

New Science in Parallel podcast episode 🎙️: Paul M. Sutter talks about his many career hats: astrophysicist, host, author, and much more. We also discuss his most recent book, Rescuing Science, which critiques the scientific enterprise, and we discuss how science might return to its most essential core values: https://scienceinparallel.org/2024/09/paul-sutter-spaceman-adventures-in-science-and-outreach/

#HPC #scicomm #DOECSGF #outreach

ICYM, the most recent Science in Parallel podcast episode 🎙️ features Rogelio Cardona-Rivera of the University of Utah discussing their research "playing games for science." Check out the full episode at https://scienceinparallel.org/2024/08/rogelio-cardona-rivera-plays-games-for-science/ #HPC

“One of the things that's really great about the #DOECSGF is it pairs you with labs that are currently in need of tools that can be deployed,” says Ariel Kellison. The Cornell University Ph.D. student studies formal methods, computational tools that can be used in a range of settings such as medical devices and military applications to make sure that software functions as expected. She’s also worked Sandia National Laboratories in California. https://deixismagazine.org/2024/08/subduing-software-surprises/ #HPC @davidbindel

Desert background and three drones flying, one in foreground and two in background. Each one has 4 propellers.

Drones at work. Defense and other life-and-death applications are only as reliable as their software. (Photo: Air Force Staff Sgt. Rachel Simones.)

New Science in Parallel #podcast episode 🎙️: Rogelio Cardona-Rivera talks about playing games for science. Learn about technical games research and its many components and applications in #AI, #storytelling platforms and #education. https://scienceinparallel.org/2024/08/rogelio-cardona-rivera-plays-games-for-science/ #HPC #DOECSGF #creativity

Client Info