Zongyi Li

Zongyi Li is a Research Fellow at Massachusetts Institute of Technology, hosted by Kaiming He. They are currently pursuing a Ph.D. in Computing and Mathematical Sciences at Caltech (2019-2025), mentored by Anima Anandkumar and Andrew Stuart.

• Ph.D. candidate: Computing and Mathematical Sciences, Caltech (2019-2025)
• B.Sc. in Computer Science and Mathematics with a Jazz minor from Washington University in St. Louis (2015-2019)

They focus on Neural Operators for learning solution operators in Partial Differential Equations (PDEs), particularly in fluid mechanics and earth science. Their work models physical simulations with chaotic behaviors and complex geometries, showing applications in weather forecasting, carbon storage, and aerodynamics simulation.

Publications emphasize resolution-invariant models, chaotic systems, and zero-shot super-resolution capabilities. Their research combines Fourier analysis, graph networks, and physics-informed loss functions to achieve state-of-the-art performance in PDE solving with up to 1000x speedup over traditional solvers.

• Fellowships:
  • Kortschak Scholarship
  • PIMCO Fellowship
  • Amazon AI4Science Fellowship
  • Nvidia Fellowship
  • MIT Novo Nordisk AI Fellowship
• Code & Open-Source:
  • Co-developer of the NeuralOperator library
  • Implementations for Fourier Neural Operators, Graph Neural Operators, and Tensorized Neural Operators
• Media Recognition:
  • Quanta Magazine
  • MIT Tech Review
  • NVIDIA Features
  • Towards Data Science