Safety is a mainstream concern as AI models interact with millions of people—but what does safety mean when AI is physically embodied? We are entering an era of general-purpose robots, powered by the same AI that transformed vision and language, yet capable of acting in the open-ended physical world. As these systems grow more capable, ensuring their safety demands moving well beyond collision-avoidance and toward a richer, context-aware understanding; knowing, for example, that cracking an egg is unsafe during transport but appropriate while cooking. This project’s central hypothesis is that nuanced safety behaviors can only emerge from equally nuanced representations of the world. We propose a unified modeling and optimization framework for physical, semantic, distributional, and privacy hazards. The ultimate goal is robots whose safety reasoning is as nuanced as our own, and whose behavior can be shaped by anyone: engineers, policymakers, and everyday users.
Awards and Achievements
- DARPA Young Faculty Award ( 2025)
- NSF CAREER Award ( 2025)
- Amazon Research Award ( 2025)
- Google Research Scholar Award ( 2024)
- Finalist for Best Paper Award of the IEEE RAS Technical Committee on Robot Control ( 2024)
- Rising Stars in EECS Award ( 2021)
- Honorable Mention for the T-RO Best Paper Award ( 2020)
- NSF Graduate Research Fellowship ( 2016)