AI People Wiki

Builders & open source

Woosuk Kwon

Co-founder and CTO, Inferact

  • Co-founder and CTO, Inferact14
  • Lead maintainer, vLLM project2

Woosuk Kwon co-created vLLM, an open-source engine for serving and running large language models, and co-leads the project; its governance document lists Kwon among the lead maintainers responsible for overall direction and strategy. Kwon is the first author of the paper that introduced vLLM and its PagedAttention approach to managing KV cache memory, and is co-founder and CTO of Inferact, a startup working on AI inference.123

At a glance

Current roles
Co-founder and CTO +1 more
Sector
Company
Education
UC Berkeley
Work listed
1
Honors
1
Sources cited
5
Profile reviewed

Career

Kwon completed a PhD in Computer Science at UC Berkeley from 2021 to 2025, advised by Ion Stoica, who is also a co-author of the vLLM paper. Kwon also worked as a Research Scientist at Google DeepMind from March 2024 and as a Member of Technical Staff at Thinking Machines Lab from May 2025.13

In November 2025 Kwon co-founded Inferact, a startup focused on AI inference, and serves as its CTO. Kwon's GitHub profile lists Inferact alongside the vllm-project organization.14

Open-source work

Kwon is first author of "Efficient Memory Management for Large Language Model Serving with PagedAttention," posted to arXiv in September 2023 and published at the 29th ACM Symposium on Operating Systems Principles (SOSP). The paper proposes PagedAttention, an attention algorithm modeled on virtual memory and paging in operating systems, and presents vLLM, a serving system built on it that cuts wasted KV cache memory and lets requests share cache.35

The vLLM repository asks users who build on the engine in research to cite that paper. The project's governance document names Kwon both as one of the Project Leads, who also form the Technical Steering Committee, and as one of the lead maintainers who share responsibility for the project's direction and strategy.52

Work and contributions

  • vLLM

    Co-created vLLM and co-leads it; first author of the PagedAttention paper that introduced the system, which the repository asks users to cite.1352

Timeline

  1. Began a PhD in Computer Science at UC Berkeley, advised by Ion Stoica.1
  2. Published the vLLM and PagedAttention paper at the ACM Symposium on Operating Systems Principles (SOSP 2023).5
  3. Posted the PagedAttention paper introducing vLLM to arXiv as first author.3
  4. Joined Google DeepMind as a Research Scientist.1
  5. Joined Thinking Machines Lab as a Member of Technical Staff.1
  6. Co-founded Inferact and became its CTO.1
  7. Completed the PhD in Computer Science at UC Berkeley.1

Other documented roles

Roles the sources document that are not confirmed as current.

Honors and recognition

  • Sequoia Open Source Fellowship · Sequoia Capital, 20241

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

5 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Woosuk KwonWoosuk Kwon · Official page · accessed Oct 21, 2026
  2. vLLM governance processvLLM project · Documentation · accessed Oct 21, 2026
  3. Efficient Memory Management for Large Language Model Serving with PagedAttentionarXiv · Paper · published Sep 12, 2023 · accessed Oct 21, 2026
  4. WoosukKwon on GitHubGitHub · Official page · accessed Oct 21, 2026
  5. vllm-project/vllm READMEvLLM project (GitHub repository) · Repository · accessed Oct 21, 2026