Researching learning-based control strategies for legged robots, with a focus on dynamic locomotion and task-driven behavior in unstructured environments. My work centers on developing reinforcement learning pipelines in simulation and translating them to hardware for robust real-world deployment.

I contribute to end-to-end system development spanning environment design, reward shaping, policy training, and evaluation. This includes integrating learned controllers with robot dynamics models, debugging instability modes, and analyzing sim-to-real gaps under latency, sensing, and contact uncertainty constraints.

My impact

• Designed and trained reinforcement learning policies for legged robotic tasks in Isaac Gym

• Built custom simulation environments and reward structures to drive stable, task-oriented locomotion

• Investigated sim-to-real transfer challenges including actuation delay, state estimation noise, and contact modeling

• Contributed to hardware validation and performance analysis of learned controllers