Pushmeet Kohli is an Indian British computer scientist and Vice President of research at Google DeepMind, where he heads the Science and Strategic Initiatives Unit. He is best known for leading the development of AlphaFold, a deep learning system that predicts protein structures, and has contributed to numerous other AI projects. In 2023, Time magazine named him one of the 100 most influential people in AI.
Kohli's research spans Machine learning, computer vision, and Artificial intelligence, with applications in computational biology, program synthesis, and optimization. His work has earned multiple best paper awards and test-of-time honors.
Education
Kohli earned a Bachelor of Technology (BTech) in Computer Science and Engineering from the National Institute of Technology, Warangal, in India. He then moved to the United Kingdom, where he completed a PhD in computer vision at Oxford Brookes University in 2007, supervised by Philip Torr. His doctoral thesis focused on discrete optimization techniques for vision problems.
Early Career
After his PhD, Kohli worked as a postdoctoral associate at the Psychometric Centre at the University of Cambridge, where he applied machine learning to psychological measurement. He then joined Microsoft Research as a partner scientist and director of research, leading projects in computer vision and crowdsourcing. During this period, he contributed to the development of the Kinect's human pose estimation system, which used depth sensors to track body movements in real time.
At Google DeepMind
Kohli joined Google DeepMind (then DeepMind Technologies) in the mid-2010s, eventually rising to Vice President of research. He leads the Science and Strategic Initiatives Unit, which focuses on applying AI to scientific discovery and addressing societal challenges. Under his supervision, DeepMind has produced several landmark systems, including AlphaFold, AlphaEvolve, SynthID, and Co-Scientist.
Major Projects
AlphaFold
AlphaFold is an AI system that predicts the 3D structure of proteins from their amino acid sequences. Kohli led the team that developed AlphaFold, which achieved breakthrough accuracy in the Critical Assessment of Protein Structure Prediction (CASP) competitions. The system has been used to predict structures for hundreds of millions of proteins, accelerating research in biology and medicine.
AlphaEvolve
AlphaEvolve is a general-purpose evolutionary coding agent that optimizes computer programs through iterative mutation and selection. It has been applied to code superoptimization, finding more efficient implementations of algorithms.
SynthID
SynthID is a system for watermarking and detecting AI-generated content, including images and text. It embeds imperceptible watermarks that can be verified later, helping to combat misinformation and ensure content authenticity.
Co-Scientist
Co-Scientist is an AI agent designed to generate and test new scientific hypotheses. It uses large language models to propose experiments and analyze results, potentially accelerating the pace of discovery.
Other Contributions
Kohli has also contributed to AlphaTensor, a reinforcement learning agent that discovers faster matrix multiplication algorithms; AlphaCode, a system for competition-level code generation; FunSearch, which uses large language models to search program spaces; and AlphaGenome and AlphaMissense, models that predict the effects of genetic mutations. His earlier work includes neural program synthesis, probabilistic programming, and behavioral analysis using online networks.
Research Themes
Kohli's research often combines discrete optimization with deep learning. He has worked on Residual Network (ResNet) architectures and Loss Functions for vision tasks, and has explored Data Augmentation techniques. His interest in psychometrics led to methods for personality prediction from social media data. In recent years, he has focused on applying AI to scientific problems, such as controlling magnetic confinement for fusion and solving the fractional electron problem using learned density functionals.
Awards and Honours
Kohli's work has been recognized with numerous awards:
- Koenderink Prize (Test of Time award) from the European Conference on Computer Vision
- British Machine Vision Association and Society for Pattern Recognition (BMVA) Sullivan Prize for best PhD thesis
- IEEE Mixed Augmented Reality (ISMAR) Impact Paper award
- Lasting Impact Award from the ACM Symposium on User Interface Software and Technology
- Best paper award at the International World Wide Web Conference 2014
- Best paper award at the European Conference on Computer Vision (ECCV) 2010
- Best paper award at the Conference on Uncertainty in Artificial Intelligence (UAI)
He was also listed in the Time 100 AI list in 2023.
Impact and Legacy
Kohli's leadership of AlphaFold has had a transformative effect on structural biology, with the system's predictions being used by researchers worldwide. His broader work on AI for science has helped establish a new paradigm where machine learning accelerates discovery in fields like genomics, materials science, and physics. As of 2025, he continues to direct research at Google DeepMind, focusing on strategic initiatives that bridge fundamental AI research and practical applications.