Koray Kavukcuoglu

Koray Kavukcuoglu is an artificial intelligence researcher and executive who serves as Senior Vice President of Google DeepMind and Chief AI Architect at Google. He is known for his work on deep learning and reinforcement learning at Google DeepMind, including contributions to research associated with Deep Q-Networks (DQN) and WaveNet, with Google crediting him for establishing DeepMind’s deep learning team.^[3][6] In August 2026, he assumed day-to-day leadership of Google DeepMind, reporting directly to Sundar Pichai and overseeing Gemini model development, frontier AI research, and the Gemini app and developer teams.^[3][4][5]

Education

Kavukcuoglu studied aerospace engineering at Middle East Technical University (METU), where he completed undergraduate and graduate studies. His 2003 master's thesis examined wing flutter analysis. He later completed a PhD in Computer Science at New York University under Yann LeCun. His dissertation, Learning Feature Hierarchies for Object Recognition, was submitted in January 2011.^[9][10][11]

Career

After completing his PhD, Kavukcuoglu worked as a Postdoctoral Research Scientist at New York University before moving to industry as a Research Scientist at NEC Laboratories America from 2011 to 2012. His earlier engineering roles included working as a Software and Control Engineer at Roketsan from September 2001 to September 2002 and an internship at Siemens in 2000.^[2]​

Kavukcuoglu joined DeepMind in 2012 as a research scientist after his time at NEC Laboratories America and later held research leadership positions including Director of Research, Vice President of Research, and, from January 2024, Chief Technology Officer of Google DeepMind.^[7][8][6] Google credits him with establishing DeepMind’s deep learning team and contributing to research associated with Deep Q-Networks (DQN), a deep reinforcement learning algorithm that learned to play Atari games directly from pixels, and WaveNet, a generative model for raw audio.^[3][6]

In June 2025, while serving as Chief Technology Officer of Google DeepMind, Kavukcuoglu was appointed Google’s first Chief AI Architect, a new senior vice president role reporting directly to CEO Sundar Pichai and focused on accelerating the integration of Gemini models into Google products.^[1][8][7] In August 2026, he became Senior Vice President of Google DeepMind and assumed its day-to-day leadership after Demis Hassabis moved to the roles of Chair of Google DeepMind and Chief Scientist of Alphabet. In this capacity, Kavukcuoglu reports directly to Pichai and oversees Gemini model development, frontier AI research, and the Gemini app and developer teams.^[3][4][5][6]

Research and Contributions

Kavukcuoglu’s research centers on deep learning, reinforcement learning, and representation learning, with an emphasis on learning hierarchical feature representations for perception tasks.^[11] His doctoral work at New York University focused on convolutional and energy-based models for object recognition, exploring how multilayer neural networks can automatically learn useful visual features from data.^[11]

At DeepMind, Google credits him with helping to establish the organization’s deep learning team and contributing to research associated with Deep Q-Networks (DQN), which demonstrated deep reinforcement learning agents that learned to play Atari 2600 games directly from pixels, and WaveNet, a generative model for raw audio that underpinned later speech synthesis systems.^[3] His publication record spans academic work from his NYU period and applied research at DeepMind, including co-authored papers in machine learning and computer vision venues that document these advances in deep learning and reinforcement learning.^[11]