PeopleofSuperintelligence

Researcher

Jordan Hoffmann

Jordan Hoffmann is an AI researcher whose work spans language-model retrieval, commonsense evaluation, and compute-efficient training. He earned a Ph.D. in applied mathematics from Harvard after researching insect wing patterns at Harvard SEAS. He coauthored RETRO and the Chinchilla training study; NeurIPS named the latter a 2022 Outstanding Paper. His research also tested language models' commonsense knowledge, and in 2025 Microsoft AI listed him among the post-training and evaluation contributors for MAI-DS-R1. Microsoft named him to lead its London AI hub in 2024; as of October 2026, he is at Microsoft AI after earlier work at DeepMind and Inflection.

Now at
Microsoft
Badges

Timeline

  1. 2018

    Modeling insect wing patterns

    In 2018, he is a Ph.D. candidate at Harvard SEAS and coauthors a mathematical model of insect wing venation. The study compares the geometry of vein-bounded areas across species.

  2. 2018–2019

    A Harvard applied math PhD

    In 2018, he is a Ph.D. candidate at Harvard SEAS; he completes his doctorate with the Rycroft Group in Fall 2019. His Ph.D. is in applied mathematics.

  3. 2021

    Retrieval-enhanced language modeling

    In 2021, he coauthors RETRO, a system that adds retrieval over a database of text passages to a transformer language model.

  4. 2022

    Testing language models for commonsense

    A 2022 study he coauthors evaluates large language models' commonsense knowledge in zero-shot and few-shot settings. It finds larger models and few-shot evaluation do not reach human-level commonsense performance.

  5. 2022

    A NeurIPS Outstanding Paper

    In 2022, he coauthors a study on compute-optimal language-model training, which finds model size and training data should scale together for a fixed compute budget. It introduces Chinchilla, a 70-billion-parameter model that outperforms Gopher at the same compute budget; NeurIPS names the paper an Outstanding Paper.

  6. April 2024

    Leading Microsoft AI London

    Microsoft names him to lead its new London AI hub in April 2024.

  7. April 2025

    MAI-DS-R1 post-training

    The April 2025 MAI-DS-R1 launch post lists him among its post-training and evaluation contributors. The open-weights model is a DeepSeek R1 variant.

  8. As of October 2026

    At Microsoft AI

    He is at Microsoft AI after working at Inflection and serving as a Research Scientist at DeepMind in London.

Career

  1. NowMicrosoftLead of the London AI hub
  2. Google DeepMindResearch Scientist

Education

  1. 2019Harvard UniversityPhD in applied mathematics

From the biography.

  1. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Jordan Hoffmann. 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. 16% of AI career moves in the 2020s went from a university to a companyIncludes Jordan Hoffmann. 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.
  3. 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 Microsoft. 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.
  4. Google had the most people shaping AI every year from 2014 to 2025Includes Microsoft. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.

Sources and credits

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

  1. Lord of the wings seas.harvard.edu. Checked 21 October 2026.
  2. Members of the Rycroft Group people.math.wisc.edu. Checked 21 October 2026.
  3. Jordan Hoffmann jhoffmann.org. Checked 21 October 2026.
  4. Improving language models by retrieving from trillions of tokens deepmind.google. Checked 21 October 2026.
  5. A Systematic Investigation of Commonsense Knowledge in Large Language Models aclanthology.org. Checked 21 October 2026.
  6. An empirical analysis of compute-optimal large language model training deepmind.google. Checked 21 October 2026.
  7. Announcing the NeurIPS 2022 Awards blog.neurips.cc. Checked 21 October 2026.
  8. Announcing new Microsoft AI Hub in London blogs.microsoft.com. Checked 21 October 2026.
  9. Introducing MAI-DS-R1 techcommunity.microsoft.com. Checked 21 October 2026.