Hantao Yu (余晗涛)
Office: Mudd 416, New York, NY, 10027
Email: hantao dot yu at columbia dot edu
[Google Scholar] [CV]
Hantao Yu (余晗涛)
Office: Mudd 416, New York, NY, 10027
Email: hantao dot yu at columbia dot edu
[Google Scholar] [CV]
Hello, welcome to my homepage!
I am a 4th year (Fall 2022 - present) Computer Science Ph.D. candidate at Columbia, advised by Josh Alman and Alexandr Andoni.
I graduated from UC San Diego in June 2022 with a B.S. in Mathematics and a minor in Computer Science.
During undergrad I worked with Chris Umans and Max Hopkins.
Research interests (please refer to my research page for detailed publication)
Most recently I have been working on Mixed-Precision Training and Asynchronous Reinforcement Learning.
I have worked on mathematical foundations of LLMs. In particular, I aim at understanding the connections between expressivity, efficiency and parallelism in Transformers from an algorithmic perspective.
I have worked as a student researcher at Google (Feb 2025 - Aug 2025), hosted by Clayton Sanford and Jean Pouget-Abadie.