CV

Education

• 2012-2016 Xiamen University, Computational Mathematics
• 2016-2022 Zhejiang University, Computational Mathematics

Experience

• 2025-Present Shanghai AI Laboratory, Researcher — Physics Simulation / Embodied AI / Real2Sim2Real
  • Developed SIM1, a physics-aligned simulator for deformable manipulation that scales 200 human demonstrations into over 10,000 synthetic trajectories; policies trained entirely on synthetic data achieve 90% zero-shot success on real robots.
  • Developed Tac2Real, a reliable multi-GPU visuotactile simulation framework with TacAlign for online reinforcement learning, enabling zero-shot Sim2Real transfer on robotic peg insertion.
  • Researched soft-robotic Sim2Real via conditional flow matching, reducing the simulation-to-reality gap on tensile and fin-ray gripping benchmarks.
• Jul 2022-2025 NetEase Fuxi (网易伏羲), Game AI / Physics Simulation Algorithm Engineer
  • Developed real-time soft-body simulation for game characters using LBS-subspace skeletons and subspace Projective Dynamics with collision, achieving over 200 FPS on CPU; an MLP accelerator reduced per-frame cost to under 1 ms.
  • Built controllable, collision-aware motion-generation methods for non-humanoid characters, including modal animation, automatic LBS rigging, and actuation-energy optimization coupled with elastic energy.
  • Implemented GPU-parallel contact solvers for interactive simulation: PNCG-IPC in Taichi and NVIDIA Warp (SIGGRAPH 2024) and MAS-PNCG in C++/CUDA, delivering over 3× speedup over prior state of the art.
  • Developed MPM-based excavator-soil/rock interaction simulation and Sim2Real techniques for intelligent excavator control.

Publications