AI Agents for Science
This class was held at the University of Chicago as CMSC 35370 in Autumn 2025.
The online curriculum provides pointers to readings and to the slides used in the class.
An agentic Scientific Discovery Platform (SDP) is an integrated environment that combines reasoning-capable AI with scientific and engineering resources—such as literature collections, simulation codes, experimental platforms, and knowledge bases—to accelerate the pace of discovery. Recent advances in large language models (LLMs) and related technologies make it possible to build such platforms that can automate key aspects of scientific work: synthesizing information from the literature, generating and prioritizing hypotheses, designing and executing protocols, running simulations or experiments, and interpreting results.
To explore and advance these ideas, we offered this intensive class in Fall 2025. The course was targeted at graduate students and advanced undergraduates eager to engage with this emerging field. The class emphasized “learning by doing”: instructors provided infrastructure, working examples, and state-of-the-art perspectives on agentic AI and scientific discovery, while students will design, build, and apply their own SDPs to tackle domain-relevant problems.
The course engaged both computer scientists, most interested in how to build such platforms, and domain scientists, most interested in how to apply them. Our aim was to create a mutually instructive environment in which diverse expertise contributes to advancing the science and practice of agentic discovery.