Perry (Peiyuan) Zhang

Hi there!

I am Peiyuan (Perry) Zhang, a second-year PhD student in Computer Science at UC San Diego, advised by Prof. Hao Zhang. My research focuses on efficient architectures for video generation, as well as post-training and inference.

Previously, I worked on LLMs and VLMs at SUTD and NTU, and interned at ByteDance Seedance. I enjoy building scalable systems and exploring new ideas at the intersection of machine learning, vision, and efficiency. I believe that great ML scientists are, fundamentally, exceptional software engineers.

News

• Nov 2025: VSA is used by Meituan LongCat Video, and STA is adopted by Hunyuan Video 1.5.
• Sep 2025: VSA was accepted at NeurIPS 2025.

Projects

• FastVideo: A unified inference and post-training framework for accelerated video generation.
• TinyLlama: An open endeavor to pretrain a 1.1B Llama model on 3 trillion tokens.
• lmms-eval: One-for-all LMMs evaluation package.
• EasyContext: LLM long-context training made simple.

See Publications for selected papers and Google Scholar for the full list.

See Experience for research and internship history.