My research interests focus on building the algorithmic foundations for generalizable autonomy, enabling robots to acquire both cognitive and dexterous skills for seamless human interaction. I am specifically interested in the use of structured inductive biases and causality for decision making to enhance robot learning. My work explores several key technical pillars:

Multi-modal Representations: Developing object-centric and spatiotemporal event representations to help robots understand complex environments.
Reinforcement Learning (RL): Advancing self-supervised pre-training for RL and control, as well as establishing principles for efficient dexterous skill acquisition.
Computer Vision: Leveraging 3D vision and foundation models to bridge the gap between perception and action.
Applied Autonomy: Implementing these theoretical foundations in high-impact domains such as personal robotics, manufacturing, and self-driving labs