AI People Wiki

Builders & open source

Koray Kavukcuoglu

SVP, Google DeepMind, Google DeepMind

Koray Kavukcuoglu is SVP of Google DeepMind and Chief AI Architect at Google, where Google's author page says Kavukcuoglu leads the development of Google DeepMind's generative AI models and their integration into Google products. In an earlier role as VP of Research at Google DeepMind, Kavukcuoglu founded the deep learning team, whose work included DQN and WaveNet. Kavukcuoglu is a co-author of the 2013 paper Playing Atari with Deep Reinforcement Learning and the 2016 WaveNet paper, and is listed among the executive sponsors or sponsors of the Gemma, Gemma 2, and Gemma 3 technical reports.1234567

At a glance

Current roles
SVP, Google DeepMind +1 more
Sector
Company
Work listed
3
Sources cited
12
Profile reviewed

Career

As VP of Research at Google DeepMind, Kavukcuoglu founded the deep learning team, whose work included DQN and WaveNet. As SVP of Google DeepMind and Chief AI Architect at Google, Kavukcuoglu leads the development of Google DeepMind's generative AI models and their integration into Google products.1

Kavukcuoglu wrote Google's October 13, 2026 announcement of Gemini 4 Argon, a frontier model for coding, enterprise knowledge work and cyber defense offered to developers, enterprises, scientists and consumers. Kavukcuoglu is also listed among the authors of the Gemma, Gemma 2 and Gemma 3 open-model reports.2567

Research

Kavukcuoglu's early papers include first-authored work on fast inference in sparse coding algorithms for object recognition (2010) and the co-authored "Natural Language Processing (almost) from Scratch" (2011).89

At DeepMind, Kavukcuoglu co-authored papers on deep reinforcement learning such as "Playing Atari with Deep Reinforcement Learning" (2013) and "Asynchronous Methods for Deep Reinforcement Learning" (2016), as well as Spatial Transformer Networks (2015), WaveNet (2016) and "Neural Discrete Representation Learning" (2017), which introduced the VQ-VAE.31011412

Work and contributions

  • Gemma Open weights

    Listed among the executive sponsors in the 2024 Gemma report, which releases pretrained and fine-tuned checkpoints, and among the sponsors of the Gemma 2 and Gemma 3 technical reports.567

  • Deep Q-Network (DQN) Published research

    Co-author of Playing Atari with Deep Reinforcement Learning (2013), which trains a convolutional network with a variant of Q-learning to play Atari 2600 games from raw pixels.31

  • WaveNet Published research

    Co-author of the 2016 WaveNet paper, which introduces an autoregressive neural network for generating raw audio waveforms and applies it to text-to-speech.41

Timeline

  1. Released "Fast Inference in Sparse Coding Algorithms with Applications to Object Recognition" as first author.8
  2. Co-authored "Natural Language Processing (almost) from Scratch".9
  3. Co-authored "Playing Atari with Deep Reinforcement Learning".3
  4. Co-authored "Spatial Transformer Networks".11
  5. Co-authored "Asynchronous Methods for Deep Reinforcement Learning".10
  6. Co-authored "WaveNet - A Generative Model for Raw Audio".4
  7. Co-authored "Neural Discrete Representation Learning", which introduced the VQ-VAE.12
  8. Listed among the authors of the first Gemma open-model report.5
  9. Announced Gemini 4 Argon as SVP of Google DeepMind and Chief AI Architect at Google.2

Other documented roles

Roles the sources document that are not confirmed as current.

Where they work

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

12 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Koray Kavukcuoglu — author page on The Keyword (Google)Google · Official page · accessed Oct 21, 2026
  2. Gemini 4 Argon: Built to solve your hardest problems (byline)Google · Announcement · published Oct 13, 2026 · accessed Oct 21, 2026
  3. Playing Atari with Deep Reinforcement Learning (arXiv 1312.5602)arXiv · Paper · published Dec 19, 2013 · accessed Oct 21, 2026
  4. WaveNet: A Generative Model for Raw Audio (arXiv 1609.03499)arXiv · Paper · published Sep 12, 2016 · accessed Oct 21, 2026
  5. Gemma: Open Models Based on Gemini Research and Technology (arXiv 2403.08295)arXiv · Paper · published Mar 13, 2024 · accessed Oct 21, 2026
  6. Gemma 2: Improving Open Language Models at a Practical Size (arXiv 2408.00118)arXiv · Paper · published Jul 31, 2024 · accessed Oct 21, 2026
  7. Gemma 3 Technical Report (arXiv 2503.19786)arXiv · Paper · published Mar 25, 2025 · accessed Oct 21, 2026
  8. Fast Inference in Sparse Coding Algorithms with Applications to Object Recognition (arXiv 1010.3467)arXiv · Paper · published Oct 18, 2010 · accessed Oct 21, 2026
  9. Natural Language Processing (almost) from Scratch (arXiv 1103.0398)arXiv · Paper · published Mar 2, 2011 · accessed Oct 21, 2026
  10. Asynchronous Methods for Deep Reinforcement Learning (arXiv 1602.01783)arXiv · Paper · published Feb 4, 2016 · accessed Oct 21, 2026
  11. Spatial Transformer Networks (arXiv 1506.02025)arXiv · Paper · published Jun 5, 2015 · accessed Oct 21, 2026
  12. Neural Discrete Representation Learning (arXiv 1711.00937)arXiv · Paper · published Nov 2, 2017 · accessed Oct 21, 2026