I am currently a Research Scientist at Meta, working on video/image generation. I received my Ph.D. in Mechanical Engineering from the University of Notre Dame in 2025. My research focuses on the intersection of Deep Learning (DL) and Physics, with a particular focus on modeling and controlling spatiotemporal dynamics governed by Partial Differential Equations (PDEs).

I believe that the fundamental differences between PDE-governed systems and tasks in Natural Language Processing (NLP) or Computer Vision (CV) necessitate the development of specialized network architectures. These systems require tailored approaches that go beyond conventional networks optimized for NLP or CV. This conviction drives my research into both DL algorithms and traditional numerical methods for PDEs, aiming to explore the unique connections and distinctions between these “trendy” networks and classical numerical techniques. Throughout my PhD, I have explored numerous strategies to incorporate physics-based priors into DL architectures, improving both performance and efficiency. The deeper I delve into this field, the more convinced I become of the potential—and the necessity—of integrating physical principles into DL. On the other hand, emerging DL methods such as diffusion models, which share deep connections with physical processes, further suggest that leveraging physics can unlock even greater potential for deep learning approaches.

Education

• Ph.D., University of Notre Dame, 2025
• B.S., Xi’an Jiaotong University, 2019