PeopleofSuperintelligence

Builder

Woosuk Kwon

Woosuk Kwon is an AI systems builder who helped build vLLM, an open-source engine for serving large language models. He earned two bachelor's degrees at Seoul National University and a computer science PhD at UC Berkeley under Ion Stoica. His work includes PagedAttention and SmartSpec, which chooses a speculation length for each request. After research work at Google DeepMind and a technical staff role at Thinking Machines Lab, he co-founded Inferact in 2025. As of October 2026, he is Inferact's CTO and co-leads vLLM.

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

Timeline

  1. 2015–2021

    Two degrees at Seoul National

    At Seoul National University, he earns B.S. degrees in computer science and engineering and mathematical sciences from March 2015 to August 2021, with a GPA of 4.18/4.30 and a rank of 1st out of 134.

  2. August 2021–December 2025

    A Berkeley computer science PhD

    At UC Berkeley's EECS Department, he starts a computer science PhD in August 2021 under advisor Ion Stoica and completes the degree in December 2025 with a thesis on vLLM and PagedAttention.

  3. 2023

    Building vLLM with PagedAttention

    In June 2023, he coauthors the vLLM launch post describing its open-source inference library built around PagedAttention and contributes to vLLM at Berkeley's Sky Computing Lab. The vLLM paper appears at SOSP 2023, and its FastChat integration supports serving Vicuna, which reaches millions of Chatbot Arena users.

  4. March 2024–May 2025

    Research scientist at Google DeepMind

    He joins Google DeepMind as a research scientist in March 2024 and works there through May 2025.

  5. June 2024

    SmartSpec speculative decoding

    In June 2024, he coauthors “Optimizing Speculative Decoding for Serving Large Language Models Using Goodput,” which introduces SmartSpec to choose a speculation length for each request.

  6. 2024

    Berkeley lead and Sequoia fellow

    In December 2024, he is one of Berkeley's graduate student team leads for vLLM as UC Berkeley and Intel announce an AI Center of Excellence to optimize generative AI inference on Intel architectures. He becomes a Sequoia Open Source Fellow in 2024.

  7. May 2025–November 2025

    Member of technical staff

    In May 2025, he joins Thinking Machines Lab as a member of technical staff and works there through November 2025.

  8. Since November 2025

    Inferact CTO and vLLM lead

    He co-founds Inferact as its CTO in November 2025 and co-leads vLLM, an open-source LLM inference engine. He is a vLLM lead maintainer and is scheduled to present “State of vLLM 2026” at the August 2026 conference as its creator and core maintainer. In September 2026, he coauthors Inferact's TPU v7 megakernel release, and as of October 2026 he is the company's co-founder and CTO.

Career

  1. Since 2025InferactCo-founder and CTO
  2. 2025Thinking Machines LabMember of technical staff
  3. 2024–2025Google DeepMindResearch scientist

Education

  1. 2025University of California, BerkeleyPh.D. in Computer Science
  2. 2021Seoul National UniversityB.S. degrees in computer science and engineering and mathematical sciences

From the biography.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Woosuk Kwon. 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.
  2. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Woosuk Kwon. 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.
  3. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Woosuk Kwon. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  4. 16% of AI career moves in the 2020s went from a university to a companyIncludes Woosuk Kwon. 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 Woosuk Kwon yet. Send us a free-licensed photo.

  1. Woosuk Kwon woosuk.me. Checked 21 October 2026.
  2. vLLM: An Efficient Inference Engine for Large Language Models | EECS at UC Berkeley www2.eecs.berkeley.edu. Checked 21 October 2026.
  3. vLLM – UC Berkeley Sky Computing Lab sky.cs.berkeley.edu. Checked 21 October 2026.
  4. vLLM: Easy, Fast, and Cheap LLM Serving with PagedAttention | vLLM Blog vllm-project.github.io. Checked 21 October 2026.
  5. Woosuk Kwon | alphaXiv alphaxiv.org. Checked 21 October 2026.
  6. [2406.14066v1] Optimizing Speculative Decoding for Serving Large Language Models Using Goodput arxiv.org. Checked 21 October 2026.
  7. UC Berkeley announces AI Center of Excellence - EECS at Berkeley eecs.berkeley.edu. Checked 21 October 2026.
  8. Sequoia Open Source Fellowship sequoiacap.com. Checked 21 October 2026.
  9. Woosuk Kwon | Cosign cosign.co. Checked 21 October 2026.
  10. Inferact — Lightspeed Venture Partners lsvp.com. Checked 21 October 2026.
  11. Governance Process - vLLM docs.vllm.ai. Checked 21 October 2026.
  12. First vLLM Conference at Ray Summit vllm.ai. Checked 21 October 2026.
  13. 700 TPS on Kimi K3: A Case for TPU Megakernels inferact.ai. Checked 21 October 2026.