Modern deep learning gives us powerful perception, but it's often data-hungry. Classical control theory gives us safety guarantees, but it struggles with the unstructured real world. I'm driven by the challenge of bridging this gap: How can we design trustworthy learning algorithms that know what they don't know? My goal is to create robotic systems that are not just smart, but fundamentally safe and robust in the face of the unknown.

We often train robots in simulators where they have perfect knowledge of the world. But when deployed, they only have noisy sensors and limited onboard compute. This Sim2Real gap isn't just about domain randomization; it's about information asymmetry. I am fascinated by how we can train policies that explicitly reason about their own perceptual limitations.