PeopleofSuperintelligence

Builder

Tom Hennigan

Also known as Thomas Hennigan

Tom Hennigan is a software engineer at Google DeepMind in London. Since 2018 he has worked there on compute infrastructure for artificial intelligence research at scale, after a senior software engineer and tech-lead post at Google from 2015 to 2018 and work on DueDil’s core distributed data-processing infrastructure in London from 2012 to 2014. In 2020 he co-authored Haiku, a neural network library for JAX, and the January 2022 Gopher report listed him on pipelining and hardware efficiency for a 280 billion parameter model. In 2022 he co-authored the NeurIPS outstanding paper on compute-optimal large language model training, and the May 2025 Gemini report named him co-lead for infrastructure.

Timeline

  1. 2012–2014

    Developer at DueDil

    From 2012 to 2014 he helps design and build DueDil’s core distributed data-processing infrastructure in London, including an in-house framework and a distributed web scraper.

  2. 2015–2018

    Tech lead at Google

    He serves as a senior software engineer and tech lead at Google from 2015 to 2018.

  3. Since 2018

    Compute infrastructure at DeepMind

    From 2018 he works as a software engineer at Google DeepMind in Greater London, on compute infrastructure that enables artificial intelligence research at scale.

  4. 2020

    Haiku for JAX

    With Trevor Cai, Tamara Norman, Lena Martens, and Igor Babuschkin, he authors Haiku in 2020, a neural network library for JAX described as Sonnet for JAX.

  5. January 2022

    Training the Gopher model

    The January 2022 Gopher report lists him on pipelining and hardware efficiency for the training infrastructure, and it presents a 280 billion parameter model called Gopher.

  6. 2022

    A NeurIPS outstanding paper

    In 2022 he co-authors the outstanding paper An empirical analysis of compute-optimal large language model training.

  7. May 2025

    Infrastructure co-lead for Gemini

    The May 2025 Gemini report names him co-lead for infrastructure and introduces a family of multimodal models across image, audio, video, and text, in Ultra, Pro, and Nano sizes.

  8. As of October 2026

    Software engineer at DeepMind

    As of October 2026, he is a software engineer at Google DeepMind in London.

Sources and credits

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

  1. Tom Hennigan | alphaXiv alphaxiv.org. Checked 16 October 2026.
  2. GitHub - google-deepmind/dm-haiku: JAX-based neural network library · GitHub github.com. Checked 16 October 2026.
  3. Scaling Language Models: Methods, Analysis & Insights from Training Gopher arxiv.org. Checked 16 October 2026.
  4. Announcing the NeurIPS 2022 Awards – NeurIPS Blog blog.neurips.cc. Checked 16 October 2026.
  5. An empirical analysis of compute-optimal large language model training neurips.cc. Checked 16 October 2026.
  6. Gemini: A Family of Highly Capable Multimodal Models arxiv.org. Checked 16 October 2026.
  7. tomhennigan (Tom Hennigan) · GitHub github.com. Checked 16 October 2026.