PeopleofSuperintelligence

Builder

Joon Kim

Joon Kim is an AI builder whose work connects enterprise language models with document parsing. He worked in machine learning at Bloomberg from 2019 to 2022, then joined Cohere's technical staff. At Cohere, he coauthored a 2023 guide to fine-tuning Command models and the 2025 Command A report, which describes a model supporting 23 languages. He coauthored Extend's RealDoc-Bench, a 2026 benchmark for document structure and field-level question answering. As of October 2026, he contributes to Extend's document-parsing work.

Now at
Extend
Badges

Timeline

  1. September 2016–May 2019

    Teaching at Maryland

    He works as an undergraduate teaching assistant at the University of Maryland from September 2016 to May 2019.

  2. June 2019–July 2022

    Machine learning at Bloomberg

    He works as a machine learning engineer at Bloomberg from June 2019 to July 2022.

  3. July 2022–January 2026

    Technical staff at Cohere

    He serves as a senior member of technical staff at Cohere from July 2022 through January 2026.

  4. June 2023

    Custom Command models

    In June 2023, he is a Member of Technical Staff at Cohere and coauthors its guide to fine-tuning custom Command models. The guide explains how to set up fine-tuning.

  5. 2025

    Command A technical report

    In 2025, he coauthors Cohere's Command A technical report. It describes an agent-optimised multilingual model that supports 23 languages of global business.

  6. May 2026

    RealDoc-Bench for agents

    In May 2026, he and three colleagues publish Extend's RealDoc-Bench article. The benchmark tests whether parsers preserve the structure that document agents need.

  7. June 2026

    Field-level document evaluation

    In June 2026, he coauthors the RealDocBench preprint, which defines a question-answering track with 1,356 field-level questions across 581 documents in four domains. It evaluates eighteen systems and reports accuracy alongside per-page cost and cache-busted latency.

  8. As of October 2026

    Document research at Extend

    As of October 2026, he contributes to Extend's document-parsing work. In September 2026, he coauthors an update to the RealDoc-Bench dataset card that adds a provenance manifest and clarifies license scope.

Career

  1. NowExtend
  2. 2022–2026CohereSenior Member of Technical Staff
  3. 2019–2022BloombergMachine Learning Engineer
  4. 2016–2019University of MarylandUndergraduate Teaching Assistant

From the biography.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Cohere. 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. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Cohere. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  3. People shaping AI work at a median of 3 organizations over a career: 3 for university starters and institute and lab starters, versus 2 for company startersIncludes University of Maryland. Across 6,853 people shaping AI with a dated work role, we estimate a median of 3 parent organizations over a career (mean 3.6), comparing role groups, starting sectors and starting decades from the 1950s to 2020–2026.
  4. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Cohere. 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.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Joon Kim's email & phone number - Founding Machine Learning Engineer at Extend | SignalHire signalhire.com. Checked 20 October 2026.
  2. How to Create Custom Command Models cohere.com. Checked 20 October 2026.
  3. Command A: An Enterprise-Ready Large Language Model arxiv.org. Checked 20 October 2026.
  4. RealDoc-Bench: A Real-World Benchmark for Document Agents extend.ai. Checked 20 October 2026.
  5. RealDocBench: A Benchmark for Field-Level QA and Layout Understanding on Real-World Regulated Documents arxiv.org. Checked 20 October 2026.
  6. RealDocBench: A Benchmark for Field-Level QA and Layout Understanding on Real-World Regulated Documents arxiv.org. Checked 20 October 2026.
  7. Add document provenance manifest and clarify license scope (#7) · Extend-AI/RealDoc-Bench at 08dde9d huggingface.co. Checked 20 October 2026.