About Me

I am currently a machine learning researcher at the model alignment team at NVIDIA. Prior to NVIDIA, I worked on recommendation systems at Amazon Advertising team.

I obtained my PhD degree at Department of Computer Science, University of Toronto, supervised by Roger Grosse. Previously, I did my undergraduate at Department of Electronic Engineering, Tsinghua University.

I am generally interested in all approaches to building strong AI models that acts intelligently, solves problems, and aids human. I am a believer of Andrew Ng’s vision that “AI is the new electricity”. I conduct research to realize this goal and enable AI for widerange positive impacts.

Specifically, my research at NVIDIA focuses on improving large language model performances in the post-training stage. My research involves model alignment algorithms, synthetic data generation, and long-horizon reason processes.

NEWS

• 2024.12, my research regarding “Knowledge Distillation in Model Alignment” goes out in NeMo-Aligner and NVIDIA Technical Blog.
• 2024.06, our model alignment toolcase, NeMo-Aligner, is accepted in COLM 2024.
• 2024.04, our model Nemotron-4-340B-Instruct (me as the leading author) is realeased (blog, HF). We also open-sourced our synthetic data generation pipeline to traing the model.