Upcoming Events

IDEaS One-Day Tutorial on AI4Science

Have you wanted to use large language models (LLMs) in your scientific research, whether in chemistry, physics, materials science, computational biology, engineering, earth and atmospheric sciences, or related areas, but felt overwhelmed by the flood of new techniques and tools? The IDEaS One-Day Tutorial on AI4Science will help you get started, offering a clear and practical introduction to LLM inference and fine-tuning so you can confidently bring these methods into your own work.

Audience
This workshop is primarily aimed at PhD students and postdocs in scientific disciplines such as physics, chemistry, materials science, computational biology, engineering, and related fields who are interested in applying LLM-based techniques to their research, including those curious about agentic workflows and LLM-assisted coding or analysis. If there is space we can also accommodate master's students.

This workshop is not intended for researchers already using advanced inference frameworks (for example, vLLM) or fine-tuning methods such as LoRA, ReFT, RLHF, RLSF, and RL with verified rewards (RLVR) in their own workflows.

Program Highlights
Hear from leading researchers and work directly with the AI for Science team:

Opening Session & Talk: Vijay Ganesh and David Sherrill, who bring decades of expertise in neuro-symbolic AI and quantum chemistry.
Talks: Alex Gagliano on AI in physics and astronomy, and Jake Rudolph on automated discovery of neutrino physics models using reinforcement learning.
Hands-on Tutorials:
- Inference & Agentic Systems for Science
- Fine-Tuning (supervised and reinforcement learning) for Science

Preparation
Please bring your own laptop and complete a few basic installations and student verifications in advance. Full instructions and setup links will be provided on the website and in follow-up emails (for registered participants). This focused one-day workshop will help you confidently start working with LLMs for science, from inference to fine-tuning, without the distraction of rapidly changing trends. Seats are limited, so please register early to secure your place.

More details will be posted on the website as we get closer to the date.