Panelist: Prof. Brett Savoie
Title: Agents as Lab Managers
Abstract: Laboratory automation will create opportunities for new experimental workflows that are data-driven and dynamically optimized. In preparation for this eventuality, we have begun developing laboratory agents that use reinforcement and active learning to navigate decision landscapes in goal-directed fashion. These agents can use information from each experiment to determine what will be useful. For example, when embedded in a digital twin of a real laboratory they can identify wasteful experiments, suggest counter-intuitive experiments with latent information, avoid redundant characterizations, and discover experiments that are lossy but nevertheless useful for avoiding a later expensive evaluation. Given the data intensity of reinforcement learning, a critical component of these case studies is the ability to use a digital twin of real laboratories for the models to understand the decision landscape. These initial case studies reveal that there are relatively few impediments to pre-training such agents and then fine-tuning them in automated real-world laboratories. Conversely, the case studies still illustrate the need for humans in the loop for goal setting, even if the means can be optimized in a data-driven manner. The automation of hypothesis generation, goal-adjustment, and troubleshooting remain beyond current capabilities.
Bio-Overview: Brett Savoie is the inaugural Coyle Mission Collegiate Professor of Engineering in the Department of Chemical and Biomolecular Engineering. Prof. Savoie graduated with degrees in chemistry and physics from Texas A&M University in 2008, obtained his Ph.D. in theoretical chemistry from Northwestern University in 2014, and from 2014-2017 was a postdoc with Thomas Miller at Caltech. In 2017, he joined the faculty of the Davidson School of Chemical Engineering at Purdue University, where he established an independent research group to develop physics-based and machine learning methods to characterize and discover new organic materials. In 2022, he was promoted to the Charles Davidson Associate Professor of Chemical Engineering at Purdue University. In July 2024, he joined the faculty at Notre Dame to advance computational materials research and lead the university’s Scientific AI initiative. Prof. Savoie is the recipient of the ACS PRF, NSF CAREER, Dreyfus Machine Learning in the Chemical Sciences, and ONR YIP awards.
The Video Recording can be accessed here.