About
Current Research
Click to switch photo.
My work connects three problems: capturing dense human-object motion, learning dexterous hand manipulation from those motion, and training policies that generalize in the real world.
I am especially interested in world models that learn how the physical world evolves, human demonstrations that provide scalable motion priors, and reinforcement learning that closes the final gap to dynamic, contact-rich control.
Research Experience
Across these roles, my focus progressed from representing human motion, to simulating physical interaction, to learning control policies that transfer to real robots.
- Peking University · Visual Computing and Learning Lab
Character animation and physics simulation on dexterous hands with Prof. Libin Liu - Stanford University
Research intern on human motion generation in the TML Lab and SVL Lab with Michelle Guo and Pei Xu, guided by Prof. Karen Liu and Prof. Jiajun Wu
Education & Awards
Education
Selected Awards
Methods & Tools
I work mainly with Python, C++, PyTorch, Isaac Lab, MuJoCo, and real-world robot systems, using Rerun and Blender for visualization and Typst and LaTeX for technical writing.
Beyond Research
I am an enthusiastic speedcuber. My official personal bests include 9.47 seconds in 3×3 at Los Angeles 2024 and 2.36 seconds in Pyraminx at Sacramento 2024. Results are available on my WCA profile.
I also write speedcubing tutorials and personal essays about places, communities, and stories that stay with me.