I'm a research lead at the Robotics and AI Institute focusing on robot learning for dexterous manipulation and locomotion". My recent work tackles whole-body and dynamic manipulation, enabling legged robots to manipulate objects at human cadence in real world environments.
- Combining Sampling and Learning for Dynamic Whole-Body Manipulation
A framework that steers a learned whole-body policy with a sampling-based reactive controller, letting legged robots dynamically manipulate large, heavy objects at human cadence.
- Sumo: Dynamic and Generalizable Whole-Body Loco-Manipulation
- Judo: A User-Friendly Open-Source Package for Sampling-Based Model Predictive Control
- Jacta: A Versatile Planner for Learning Dexterous and Whole-body Manipulation
- Dojo: A Differentiable Physics Engine for Robotics
- Fast Contact-Implicit Model-Predictive Control
- ALGAMES: a fast augmented Lagrangian solver for constrained dynamic games
| 🛠️ project name | 📖 brief description | ⭐ stars |
|---|---|---|
| Judo | hackable sampling-based MPC toolbox |  |
| Dojo | differentiable physics engine for robotics |  |
| ContactImplicitMPC | predictive control algorithm for robots that make and break contact |  |
| Algames | game-theoretic solver |  |
- PhD candidate in robotics and optimization at Stanford University working with Zachary Manchester and Mac Schwager
- research scientist intern at Google Robotics
- software engineering intern at Aurora Innovation
- undergraduate student at Ecole Centrale Paris