News

Our Center Aligns with Genesis Mission Goals

The White House has launched the Genesis Mission, “a dedicated, coordinated national effort to unleash a new age of AI‑accelerated innovation and discovery that can solve the most challenging problems of this century”. Genesis formally recognizes scientific foundation models and AI agents as core national capabilities; it will develop a platform “to train scientific foundation models and create AI agents to test new hypotheses, automate research workflows, and accelerate scientific breakthroughs”. The program:
Read more »

Foundation Models for Discovery and Exploration in Chemical Space

A paper about our molecular foundation model work MIST is now on arXiv! MIST is a family of models that spans small molecules, organometallics, mixtures/formulations, and even isotopes, relying on our Smirk tokenizer to incorporate stereochemical, isotopic, and electronic information. Using MIST, we screen over 8 million candidate electrolyte solvents in eight hours, identifying 63 Pareto-optimal candidates, and achieve state-of-the-art accuracy for mixture properties across temperature and composition, surpassing our previous GNN-based DiffMix models.
Read more »

Building AI foundation models to accelerate the discovery of new battery materials

Our work using foundation models to discover new battery materials was featured by Argonne Leadership Computing Facility (ALCF). Read their article here.

SciFM25: Scientific Discovery in the Age of AI

SciFM25: Scientific Discovery in the Age of AI is the 2nd conference on foundation models and AI agents for science, following the highly successful SciFM24 conference. The event will feature engaging talks and panel discussions from thought leaders and world-renowned experts from national labs, academia, and leading AI companies; it will also include hands-on tutorials and interactive workshops.

Summer School 2024

The SciFM Summer School 2024 was help from July 8 to July 26. The goals of the summer school and hackathon focused on developing technical plans for specific instances for SciFMs, jumpstarting the construction of SciFMs and training researchers and students to further the vision of a national SciFM ecosystem. Video recordings of lectures and tutorials from the summer school are now available on YouTube. For more information about the programming click here.

MICDE Special Issue Dedicated to SciFM24 Conference

Read coverage on the MICDE Conference for Scientific Foundation Models, including Prof. Viswanthan’s talk on ``Scaling up Materials Foundation Models’’ here.

2024 MICDE Anual Symposium: Scientific Foundation Models

Join us for the 2024 MICDE Annual Symposium at the University of Michigan on April 2nd and 3rd 2024! We’ve got an exciting lineup of the foremost experts in Scientific Foundation Models who’ll be sharing their work and discussing what’s next for the field.
Read more »

2024 MICDE Conference: Scientific Foundation Models

For the first time ever, the 2024 MICDE Symposium will focus on Scientific Foundation Models! The Symposium will bring together leaders in the field to share their perspective and lead workshops on this exciting and nascent research area. The event is free and open to everyone, but you can register to get the latest information.
Read more »

INCITE 2024 Award: Foundation Models for Molecular Design for Energy Storage and Conversion

We’ve been selected for the 2024 INCITE Award and have been awarded 200,000 node-hours on Polaris at the Argonne National Laboratory! Using this grant we’ll train a BERT-based LLM on 49B molecules to reach quantum-accuracy for molecular property prediction.
Read more »

Microsoft Research's Accelerating Foundation Models Research Initiative

We’re thrilled to announce Microsoft Research’s support for our work Leveraging In-Context-Learning of Foundation Models for Accelerated Battery Materials Design during Phase II of their initiative. As part of this award, we’ve received credits on the Azure Cloud and Azure OpenAI, which we’re using to probe the use of foundation models for battery material design.
Read more »

2023 MICDE Catalyst Grant

We’ve been awarded the 2023 Catalyst Grant by Michigan Institute for Computational Discovery & Engineering. For this grant, MICDE looks to fund “innovative research projects in computational science that combine elements of mathematics, computer science, and cyber-infrastructure.” Thanks to this grant we’re developing scaling laws for training LLMs for Molecular Discovery using Wafer-Scale Computing.
Read more »