Research
I'm a full-stack roboticist. I believe robust physical intelligence that works alongside humans won't come from scaling data quantity alone: it also takes the right hardware, the right sensors, and learning algorithms that make the most of both. My research spans:
- Policy learning for dexterity: imitation and reinforcement learning from multimodal sensing [, , , , , ].
- Scalable hardware for manipulation: designing manipulators [, ] and multimodal tactile sensors [, ] that make contact-rich tasks easier to learn, teleoperation systems for collecting high-quality data [, ], and methods that co-optimize a robot's body with its policy [, , ].
- Human–robot interaction: when [, ] and how [] humans should intervene during robot deployment, and how to continually improve policy performance from these interventions [].
