Ryan Hardesty Lewis

Ryan Hardesty Lewis is a graduate student in social and engineering systems, where he develops world models for agents operating in dynamic environments, with applications spanning transportation, cities, and other large-scale socio-technical systems. He studies how AI systems can learn to simulate how people, agents, and environments interact, then uses those simulations for planning, evaluation, and decision-making before interventions are deployed. His work brings together model-based reinforcement learning, pretraining for motion and interaction models, digital twins, and counterfactual simulation: for example, editing roadway geometry and testing how driving agents respond as the environment changes. As a MathWorks Fellow, Ryan will advance computational tools for validating these models, stress-testing policies across many scenarios, and identifying when simulated behavior is reliable, brittle, or unfair. His research aims to make simulation a trustworthy foundation for safer, more equitable infrastructure and for AI systems that reason about the physical and social world.