Archana Raja, Elizabeth Nowadnick
We propose a workshop on agentic AI for managing computational materials science workflows. Many materials simulation approaches, such as density functional theory (DFT), require practitioners to manage complex multi-step workflows that require expertise in physics, computational science, and high-performance computing. The advent of large language model (LLM)-based agents in the last few years presents a new opportunity to orchestrate and automate these workflows. This workshop will introduce users to the design of state-of-the-art multi-agent AI systems for scientific workflow automation, and how this can be applied to solve problems across materials science. It will include a hands-on tutorial on the use of TritonDFT, a recently developed agentic AI platform that automates Quantum Espresso calculations.
Friday, August 21
Symposium Location: Mariposa Room
Symposium Schedule:
1:25 pm
Welcoming Remarks
Archana Raja and Beth Nowadnick
1:30 – 1:55 pm
Title TBD
Maria Chan, Argonne National Laboratory
1:55 – 2:20 pm
A general machine learning framework for many-body interactions in real materials
Diana Qiu, Yale University
2:20 – 2:45 pm
Title TBD
Yingheng Tang, Lawrence Berkeley National Laboratory
2:45 – 2:55 pm
Introduction to AI agents for automating Density Functional Theory workflows
Beth Nowadnick, University of California, Merced
2:55 – 3:20 pm
Break
3:20 – 3:50 pm
Introduction to TritonDFT, a multi-agent AI framework for DFT automation
Kuntal Talit and Zhengding Hu, University of California, Merced and University of California, San Diego
3:50 pm
Hands-On TritonDFT Tutorial
Kuntal Talit and Zhengding Hu, University of California, Merced and University of California, San Diego