Jiaming Ji (吉嘉铭)

Co-founder & CEO, Physis

World Models
Embodied AI & Physical Simulation
Model-Based Reinforcement Learning

Email: jiamg.ji at gmail dot com or jiamg.ji at stu.pku.edu.cn

[Google Scholar][GitHub]

I am the Co-founder & CEO of Physis, a world model startup building general-purpose foundation model for physical intelligence. I am also a PhD student at the Institute of Artificial Intelligence, Peking University. My work connects world models, model-based reinforcement learning, multimodal representation learning, and post-training alignment, with the goal of enabling AI systems to predict, plan, and act in complex real-world environments.

My work has been recognized by Peking University Student of the Year (2025; 10 students university-wide), the Tencent Project Up Scholarship (2026; 15 recipients nationwide), Apple Scholars in AI/ML (2025; 2 in Mainland China), the Ant Group Intech Scholarship (2025; top 10 globally), the NSFC Grant for Young Doctoral Students (2023; sole recipient from Peking University's Intelligence Science field), the Peking University President Scholarship, the CIE-Tencent Doctoral Research Incentive Project, and ACL 2025 Best Paper. Earlier work in constrained reinforcement learning and robotic dexterous manipulation, including winning the NeurIPS 2022 MyoChallenge, continues to shape my approach to world models for planning and control.