PeopleofSuperintelligence

Builder

Zhuohan Li

Zhuohan Li is an AI systems builder whose work includes vLLM, an open-source engine for serving large language models with PagedAttention. He earned a computer science B.S. at Peking University in 2019 and a Ph.D. at UC Berkeley in 2024 under Ion Stoica, and he worked as a research intern at Google Brain and Google DeepMind from May 2021 to April 2024. He coauthored a 2024 OSDI paper on fair LLM serving and a 2025 memory-allocation framework for heterogeneous LLM embeddings called Jenga. At OpenAI from June 2024 to August 2025, he led infrastructure and inference work for gpt-oss and built training and inference systems for reasoning models; since August 2025, he is an AI Research Scientist at Meta, focusing on inference.

Now at
Meta
Badges
Links
Website, Google Scholar, X, LinkedIn and GitHub

Timeline

  1. 2015–2019

    Computer science at Peking University

    He earns a B.S. in computer science at Peking University in 2019.

  2. 2019–2024

    A Berkeley PhD

    At UC Berkeley, he completes a Ph.D. in computer science in 2024 under Ion Stoica.

  3. May 2021–April 2024

    Google Brain research internship

    From May 2021 to April 2024, he works as a research intern at Google Brain and Google DeepMind, hosted by Yanping Huang, Yuanzhong Xu, and Zhifeng Chen.

  4. 2023

    vLLM and PagedAttention

    In 2023, he coauthors the vLLM paper and helps develop its open-source LLM serving engine. PagedAttention is inspired by virtual-memory paging and manages the key-value cache; vLLM improves throughput by 2–4× at the same latency.

  5. 2024

    Fairness in LLM serving

    At OSDI 2024, he coauthors “Fairness in Serving Large Language Models,” which proposes the Virtual Token Counter, a fair scheduler based on continuous batching.

  6. June 2024–August 2025

    Inference systems at OpenAI

    From June 2024 to August 2025, he serves as a Member of Technical Staff at OpenAI. He leads infrastructure and inference work for gpt-oss and builds training and inference systems for reasoning models.

  7. 2025

    Jenga for heterogeneous LLMs

    In 2025, he coauthors Jenga, a memory-allocation framework for heterogeneous LLM embeddings. The system improves GPU memory utilization by up to 79.6% and serving throughput by up to 4.92×, with an average gain of 1.80×.

  8. Since August 2025

    AI Research Scientist at Meta

    He joins Meta as an AI Research Scientist in August 2025 and focuses on inference.

Career

  1. Since 2025MetaAI Research Scientist
  2. 2024–2025OpenAIMember of Technical Staff
  3. 2021–2024Google BrainResearch Intern
  4. 2021–2024Google DeepMindResearch Intern

Education

  1. 2024University of California, BerkeleyPh.D.
  2. 2019Peking UniversityB.S.

From the biography.

  1. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Zhuohan Li. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  2. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Zhuohan Li. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.
  3. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Zhuohan Li. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  4. 16% of AI career moves in the 2020s went from a university to a companyIncludes Zhuohan Li. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Zhuohan Li yet. Send us a free-licensed photo.

  1. Zhuohan Li zhuohan.li. Checked 21 October 2026.
  2. vLLM – UC Berkeley Sky Computing Lab sky.cs.berkeley.edu. Checked 21 October 2026.
  3. vLLM: Easy, Fast, and Cheap LLM Serving with PagedAttention | vLLM Blog vllm-project.github.io. Checked 21 October 2026.
  4. Efficient Memory Management for Large Language Model Serving with PagedAttention arxiv.org. Checked 21 October 2026.
  5. Fairness in Serving Large Language Models | USENIX usenix.org. Checked 21 October 2026.
  6. Jenga: Effective Memory Management for Serving LLM with Heterogeneity arxiv.org. Checked 21 October 2026.