Hao Zhang
UC San Diego assistant professor who leads the Hao AI Lab; FastVideo, vLLM and Chatbot Arena co-author
- Assistant Professor, Halıcıoğlu Data Science Institute and Computer Science and Engineering; leads the Hao AI Lab, University of California San Diego123
Hao Zhang is an Assistant Professor in the Halıcıoğlu Data Science Institute and the Department of Computer Science and Engineering at UC San Diego, where Zhang leads the Hao AI Lab. Zhang's research sits at the intersection of machine learning and systems, and the lab maintains FastVideo, a framework for accelerating video diffusion models. Zhang is a co-author of the vLLM (PagedAttention), DistServe and Chatbot Arena papers and co-founded the nonprofit LMSYS Org.12346
Career
Zhang completed a PhD in computer science at Carnegie Mellon University with Eric Xing from 2014 to 2020, taking leave from 2016 to 2021 to work at the machine learning platform company Petuum as Director of Scalable Machine Learning. Zhang was then a postdoctoral researcher at UC Berkeley with Ion Stoica from 2021 to 2023.21
Zhang joined UC San Diego as an assistant professor in 2023 and leads the Hao AI Lab there. From 2023 to 2026 Zhang also worked as a software engineer at Snowflake, and Zhang co-founded the nonprofit LMSYS Org, which maintains open models, evaluation platforms and systems.123
Research
Zhang's work covers LLM inference and serving systems, efficient model architectures and algorithms, open data, models and evaluations, and model-parallel training systems. Projects listed on Zhang's site include vLLM, DistServe, Lookahead Decoding, Alpa, Vicuna, FastChat and Chatbot Arena.1678
For video generation, Zhang's group published Sliding Tile Attention (2025) and Video Sparse Attention (2025), and maintains FastVideo, a framework for post-training and accelerated inference of video diffusion models whose V1 release was announced in April 2025.91045
Work
FastVideoOpen-source code
Leads the Hao AI Lab that maintains FastVideo and is the senior author of the Sliding Tile Attention and Video Sparse Attention papers that the FastVideo README asks users to cite.41910
vLLM (PagedAttention)Published research
Co-author of "Efficient Memory Management for Large Language Model Serving with PagedAttention" (2023), the paper behind vLLM, which Zhang's site lists among the lab's open-source work.61
Chatbot Arena (external site: arxiv.org)Published research
Co-author of "Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference" (2024).81
Career timeline
2014
Began a PhD in computer science at Carnegie Mellon University with Eric Xing.2
2016
Joined Petuum as Director of Scalable Machine Learning, a role held until 2021.12
2021
Began postdoctoral research at UC Berkeley with Ion Stoica.2
2023
Joined UC San Diego as an assistant professor.1
Sep 2023
Co-authored the PagedAttention paper describing the vLLM serving system.6
Jan 2024
Co-authored DistServe, on disaggregating prefill and decoding for LLM serving.7
Mar 2024
Co-authored the Chatbot Arena paper on evaluating LLMs by human preference.8
Feb 2025
Co-authored Sliding Tile Attention for fast video generation.9
Apr 24, 2025
May 2025
Co-authored VSA, a trainable sparse attention method for video diffusion.10
Other documented roles
Roles the sources document but that are not confirmed as current.
Honors and recognition
- OSDI Jay Lepreau Best Paper Award · USENIX Symposium on Operating Systems Design and Implementation (OSDI), 202121
- Google ML and Systems Junior Faculty Award · Google, 202521
- Sloan Research Fellowship · Alfred P. Sloan Foundation, 202621
- NSF CAREER Award · U.S. National Science Foundation, 202621
- EuroSys Test-of-Time Award · EuroSys conference, 202621
Related people
More sourced profiles at the same organization, in the same field, or from the same school.
- Alexei A. EfrosAlso at University of California, Berkeley
- Byron HsuTraining systems
- Adam GleaveAlso from UC Berkeley
- Ali GhodsiAlso at University of California, Berkeley
- Daniel HanTraining systems
- Albert GuAlso from Carnegie Mellon
- Anca DraganAlso at University of California, Berkeley
- Furu WeiInference and serving
- Alex SmolaAlso from UC Berkeley
- Andrew WatermanAlso at University of California, Berkeley
- Ion StoicaAlso at University of California, Berkeley
- Ali FarhadiAlso from Carnegie Mellon
Is this you?
Profiles cover public professional work only, and every line cites a source. If something is out of date, missing or wrong, send a correction with a link to an official page and we will update it. Suggest a correction
Add this badge to your website or GitHub profile README. It links back to this page. It is not an endorsement or a rating.
Markdown (for a README)
[](https://unitedstatesofamericasuperintelligence.com/people/hao-zhang/)
HTML
<a href="https://unitedstatesofamericasuperintelligence.com/people/hao-zhang/"><img src="https://unitedstatesofamericasuperintelligence.com/badges/profile.svg" alt="Hao Zhang — Profile on USASI" height="20"></a>
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.