My research interests are in creating scalable, robust, and adaptable decision-making algorithms. Towards this goal, I have worked on generative models and in-context learning, deep reinforcement and imitation learning (particularly from large offline datasets), and robust deep learning.

My current aim is to create embodied foundation models for decision-making that generalize to out-of-distribution tasks and environments via in-context learning. I believe that semi-parametric methods (e.g., retrieval + transformers) will help achieve this goal.