I build evaluation infrastructure for AI agents at Tencent Hunyuan, focusing on agent observability and trajectory analysis.

I dropped out of my PhD at Tsinghua University, where I worked on reinforcement learning and interactive robotics with Professor Yanan Sui. I received my bachelor's degree in Mechanical Engineering from Tongji University in 2024, advised by Professor Qirong Tang.

Previously, I worked as a full-stack robotics engineer, from mechanical design and manufacturing to electrical control and training AI models for robots.

Feel free to reach out to discuss AI, robotics, or potential collaborations.

MS-Emulator combines GPU-parallel simulation and reinforcement learning to reproduce whole-body musculoskeletal motion and analyze internal dynamics.

SMS-Human integrates a whole-body musculoskeletal model with visual, vestibular, proprioceptive, and tactile inputs for sensorimotor control.

MPC2 is a hierarchical model-based planning framework for high-dimensional musculoskeletal systems. I developed the reinforcement learning baseline for this work.

The best way to learn something is to build something with AI. Motivated by Hunter Zhang's work, this knowledge base contains a bottom-up learning path I customized using AI for mastering AI Infrastructure:

📚 Explore the Knowledge Base →

I regularly share AI and robotics insights on Chinese Social Media Platforms, covering topics including:

• AI research insights and practical applications
• Latest developments in Reinforcement Learning
• Robotics technology trends and breakthroughs
• Personal research journey and academic experiences

I build evaluation infrastructure for AI agents at Tencent Hunyuan, focusing on agent observability and trajectory analysis. I also share what I learn about AI and robotics.

Rednote      Douyin      Bilibili