Researcher

Jake Bruce

Also known as Jacob Bruce

Jake Bruce is a research scientist at Google DeepMind in London who works on world models, and he is the first author of Genie, which won a Best Paper award at ICML 2024. After a bachelor's degree at Acadia University, he studied for an MSc at Simon Fraser University, where he helped release the SFU Mountain Dataset of woodland robot recordings in 2015. His PhD at the Queensland University of Technology, completed in 2020, taught robots to navigate in a city from a single prior traversal, and in 2018 he coauthored the Room-to-Room benchmark for natural-language navigation in real buildings. In 2022 he coauthored Gato, a single network that plays Atari, chats and stacks blocks with a real robot arm. Genie, an 11-billion-parameter world model, learned action-controllable worlds from unlabelled internet videos in 2024, and he went on to contribute to Genie 2 in December 2024.

Timeline

  1. 2015

    The SFU Mountain Dataset

    After a computer science bachelor's degree at Acadia University in Wolfville, he studies for an MSc in computer science at Simon Fraser University in Burnaby. In 2015, with Jens Wawerla and Richard Vaughan of SFU's Autonomy Lab, he releases the SFU Mountain Dataset, robot sensor recordings from seven traversals of 4 km of woodland trails on Burnaby Mountain.

  2. 2017–2018

    One-shot robot navigation

    In 2017, with Niko Sünderhauf, Piotr Mirowski, Raia Hadsell and Michael Milford, he presents one-shot reinforcement learning for robot navigation, built on an interactive world model from a single traversal of the environment. A 2018 Conference on Robot Learning paper extends the method to more than 2 km of travel through buildings and outdoor regions, and the learned policy runs on a real robot without fine tuning.

  3. 2018

    The Room-to-Room benchmark

    He coauthors a 2018 CVPR paper with Peter Anderson, Qi Wu and others that presents the Matterport3D Simulator and the Room-to-Room (R2R) dataset, the first benchmark for visually grounded natural-language navigation in real buildings.

  4. 2020

    A PhD in robot navigation

    He completes a PhD by publication at the Queensland University of Technology in 2020, supervised by Michael Milford and Niko Suenderhauf. His thesis, Learning from limited experience, combines adaptive sequence-based localization with model-based deep reinforcement learning so that robots can localize and navigate in a city from a single prior traversal.

  5. 2022

    Gato, a generalist agent

    In 2022, he coauthors the report on Gato with Scott Reed, Konrad Zolna, Nando de Freitas and others. The same network with the same weights plays Atari, captions images, chats and stacks blocks with a real robot arm.

  6. 2024

    Genie and an ICML best paper

    He is first author of Genie, posted in February 2024, an 11-billion-parameter foundation world model trained on unlabelled internet videos that generates action-controllable virtual worlds from text, images, photographs and sketches. The paper wins a Best Paper award at ICML 2024.

  7. December 2024

    Contributing to Genie 2

    In December 2024, Google DeepMind introduces Genie 2, a foundation world model that generates playable 3D environments from a single prompt image for training and evaluating embodied agents. Led by Jack Parker-Holder, the project counts him among its key contributors.

  8. As of October 2026

    Research scientist at Google DeepMind

    He is a research scientist at Google DeepMind in London. In 2025, he gives additional support to Genie 3, a world model from the same research line.

Career

  1. NowGoogle DeepMindResearch scientist

Education

  1. 2020Queensland University of TechnologyPhD
  2. Acadia UniversityBachelor's degree
  3. Simon Fraser UniversityMSc student

Sources and credits

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

  1. Jake Bruce jakebruce.ca. Checked 21 October 2026.
  2. SFU Mountain Dataset autonomy.cs.sfu.ca. Checked 21 October 2026.
  3. One-Shot Reinforcement Learning for Robot Navigation with Interactive Replay arxiv.org. Checked 21 October 2026.
  4. Learning Deployable Navigation Policies at Kilometer Scale from a Single Traversal arxiv.org. Checked 21 October 2026.
  5. Vision-and-Language Navigation: Interpreting visually-grounded navigation instructions in real environments arxiv.org. Checked 21 October 2026.
  6. Learning from limited experience: Real-world robot navigation from a single traversal | QUT ePrints eprints.qut.edu.au. Checked 21 October 2026.
  7. A Generalist Agent arxiv.org. Checked 21 October 2026.
  8. Genie: Generative Interactive Environments arxiv.org. Checked 21 October 2026.
  9. ICML 2024 Awards icml.cc. Checked 21 October 2026.
  10. Genie 2: A large-scale foundation world model — Google DeepMind deepmind.google. Checked 21 October 2026.