PeopleofSuperintelligence

Lab leader

Koray Kavukcuoglu

Koray Kavukcuoglu is a machine-learning researcher and Google DeepMind executive whose work spans computer vision, deep reinforcement learning, and generative AI. He presents visual-feature research at NYU in 2010 and later coauthors the DQN paper, Pixel Recurrent Neural Networks, and Asynchronous Methods for Deep Reinforcement Learning. Pixel Recurrent Neural Networks receives an ICML 2016 Best Paper Award; he is DeepMind's VP of Research by December 2019 and joins the Royal Academy of Engineering's Fellowship in 2022. As of October 2026, he is Google DeepMind's senior vice president and Google's Chief AI Architect; ICML has also given his asynchronous-learning paper the 2026 Test of Time Award.

Works
119
Citations
156,513
h-index
67

Publication numbers from OpenAlex, as of 14 October 2026.

Photo of Koray Kavukcuoglu

Timeline

  1. 2010

    Learning visual features

    At an NYU graduate-student and postdoctoral seminar, he presents unsupervised feature-learning methods and a hierarchical system for object recognition.

  2. 2013

    Teaching an Atari agent

    With colleagues, he coauthors Playing Atari with Deep Reinforcement Learning, which learns control policies from high-dimensional sensory input through reinforcement learning.

  3. 2016

    Drawing images pixel by pixel

    He coauthors Pixel Recurrent Neural Networks, an image model that predicts pixels in sequence with two-dimensional recurrent layers. Google Research lists it as a recipient of an ICML 2016 Best Paper Award.

  4. 2016

    Parallel learning for control

    He coauthors Asynchronous Methods for Deep Reinforcement Learning, which uses asynchronous gradient descent for deep neural network controllers and reports that parallel actor-learners stabilize training.

  5. December 2019

    Building a research group

    In December 2019, he serves as DeepMind's VP of Research. He founded DeepMind's deep-learning team, which worked on DQN and WaveNet.

  6. 2022

    Joining the academy

    He is among the 72 Fellows elected by the Royal Academy of Engineering in 2022.

  7. July 2026

    A test of time

    ICML gives its 2026 Test of Time Award to Asynchronous Methods for Deep Reinforcement Learning, which he coauthored with seven colleagues.

  8. As of October 2026

    Scaling generative AI

    He serves as senior vice president of Google DeepMind and Chief AI Architect at Google. He leads development of the lab's generative AI models and their integration into Google products.

Sources and credits

Photo: blog.google; Koray Kavukcuoglu; creator not stated. License: Not stated by source. Source: blog.google. Cropped to a square and resized. Ask us to remove or replace this photo.

  1. Learning Hierarchies of Features for Object Detection | Department of Mathematics | NYU Courant math.nyu.edu. Checked 14 October 2026.
  2. Playing Atari with Deep Reinforcement Learning arxiv.org. Checked 14 October 2026.
  3. Pixel Recurrent Neural Networks proceedings.mlr.press. Checked 14 October 2026.
  4. ICML 2016 & Research at Google research.google. Checked 14 October 2026.
  5. Asynchronous Methods for Deep Reinforcement Learning proceedings.mlr.press. Checked 14 October 2026.
  6. From unlikely start-up to major scientific organisation: Entering our tenth year at DeepMind deepmind.google. Checked 14 October 2026.
  7. Koray Kavukcuoglu blog.google. Checked 14 October 2026.
  8. Royal Academy of Engineering welcomes 72 new Fellows raeng.org.uk. Checked 14 October 2026.
  9. Announcing the ICML 2026 Awards – ICML Blog blog.icml.cc. Checked 14 October 2026.