Zoubin Ghahramani
Last reviewed
Sources
7 citations
Review status
Source-backed
Revision
v1 · 1,696 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
7 citations
Review status
Source-backed
Revision
v1 · 1,696 words
Add missing citations, update stale details, or suggest a clearer explanation.
Zoubin Ghahramani (born 8 February 1970) is a British-Iranian computer scientist and one of the leading figures in Bayesian machine learning and probabilistic approaches to artificial intelligence. He is Vice President of Research at Google DeepMind and Professor of Information Engineering at the University of Cambridge, two posts he holds concurrently. [1][2][3] From 2021 until its 2023 merger into Google DeepMind he was head of Google Brain, the research group that drove much of Google's work on generative AI, and he is a co-author of the technical report behind the Gemini 2.5 model family. [4][5]
Before joining Google he was Chief Scientist of Uber, a position he reached after the ride-hailing company acquired the startup Geometric Intelligence, which he had co-founded with Gary Marcus and others. [2][6] Across a research career spanning machine learning, statistics and computational neuroscience, he has published roughly 300 papers that have together been cited more than 125,000 times. [5] He was elected a Fellow of the Royal Society in 2015 and received the Society's Milner Award in 2021 for his contributions to probabilistic machine learning. [3]
According to his Wikipedia biography, Ghahramani was born in Moscow to an Iranian family with roots in Shiraz, and part of his schooling took place at the American School of Madrid in Spain. [1] He went on to the University of Pennsylvania, where he completed a joint undergraduate degree in computer science and cognitive science in 1990. [1][2]
He pursued doctoral study at the Massachusetts Institute of Technology in the Department of Brain and Cognitive Sciences, earning his PhD in 1995 with a dissertation titled "Computation and Psychophysics of Sensorimotor Integration." His advisor was Michael I. Jordan, and his Wikipedia biography also lists the computational neuroscientist Tomaso Poggio among his supervisors. [1] During this period he and Daniel Wolpert, working with Jordan, produced an influential 1995 paper in Science on an internal forward model for sensorimotor control, an early example of the probabilistic modeling that would define his later work. [5]
After MIT he moved to the University of Toronto as a postdoctoral fellow, joining the group of Geoffrey Hinton, one of the founders of modern neural network research. [1]
In 1998 Ghahramani became one of the founding faculty members of the Gatsby Computational Neuroscience Unit at University College London, where he remained until 2005. [1][4] He held a concurrent appointment as Associate Research Professor in the Machine Learning Department at Carnegie Mellon University, a position he kept for more than a decade from 2003 to 2012. [1][4]
He joined the University of Cambridge in 2006 as Professor of Information Engineering in the Department of Engineering, where he leads the Machine Learning Group, and he has held the chair ever since. [4] He became a Fellow of St John's College, Cambridge in 2009 and was elected an Honorary Fellow of the college in 2025. [1] In Cambridge he also took on a series of institution-building roles: he served as the founding Cambridge Director of the Alan Turing Institute, the United Kingdom's national institute for data science and AI, and as a founding deputy director of the Leverhulme Centre for the Future of Intelligence. [2]
| Period | Role | Organization |
|---|---|---|
| 1990 | Joint BA, computer science and cognitive science | University of Pennsylvania |
| 1995 | PhD, brain and cognitive sciences | Massachusetts Institute of Technology |
| 1995 to 1998 | Postdoctoral fellow (with Geoffrey Hinton) | University of Toronto |
| 1998 to 2005 | Founding faculty member | Gatsby Unit, University College London |
| 2003 to 2012 | Associate Research Professor (concurrent) | Carnegie Mellon University |
| 2006 to present | Professor of Information Engineering | University of Cambridge |
| 2016 to 2020 | Chief Scientist and VP for AI | Uber |
| 2020 to 2023 | Senior Research Director, then VP of Research and head of Google Brain | |
| 2023 to present | Vice President of Research | Google DeepMind |
Ghahramani is best known for advancing probabilistic and Bayesian methods as a unifying framework for machine learning, the idea that a learning system should reason explicitly about uncertainty using the rules of probability. The Royal Society credits him with pioneering work on semi-supervised and active learning, on sparse Gaussian processes, and on infinite-dimensional nonparametric models such as the infinite latent feature model. [3]
Several strands of this work have become standard references in the field:
He set out his broader vision in a 2015 review article in Nature, "Probabilistic machine learning and artificial intelligence," which argued for treating uncertainty as central to intelligent systems and surveyed probabilistic programming, Bayesian optimization and automatic model discovery. [5] That theme also motivated his "Automatic Statistician" project, an effort to build systems that can search over statistical models and explain their findings in plain language.
The table below lists a selection of his most cited works, drawn from his Google Scholar profile, where his publications have accumulated more than 125,000 citations with an h-index above 130. [5]
| Year | Work | Note |
|---|---|---|
| 1995 | "An internal model for sensorimotor integration" (Science) | with Daniel Wolpert and Michael I. Jordan |
| 1999 | "An Introduction to Variational Methods for Graphical Models" | foundational review of variational inference |
| 2003 | "Semi-Supervised Learning Using Gaussian Fields and Harmonic Functions" | with Xiaojin Zhu and John Lafferty |
| 2015 | "Probabilistic machine learning and artificial intelligence" (Nature) | programmatic review article |
| 2016 | "Dropout as a Bayesian Approximation" | with Yarin Gal; uncertainty in deep learning |
| 2025 | Gemini 2.5 technical report | Google DeepMind, large multimodal models |
In 2014 Ghahramani co-founded the machine learning startup Geometric Intelligence together with the cognitive scientist Gary Marcus, the computer scientist Kenneth Stanley and the NYU graduate Douglas Bemis. [6] The company pursued data-efficient approaches to machine intelligence inspired by cognitive science. In December 2016 it was acquired by Uber and became the nucleus of Uber AI Labs. [6] Ghahramani served as Uber's Chief Scientist and Vice President for AI from 2016 to 2020, building out the company's central machine learning research. [2]
In 2020 Ghahramani left Uber to join Google Research as a Senior Research Director and part of the leadership of the Google Brain team, while continuing to hold his Cambridge chair. [7] He was promoted to Vice President of Research in 2021 and became head of Google Brain, the group whose work underpinned products such as the Bard chatbot and the company's broader response to large language models. [1][2]
In April 2023 Google merged Google Brain with DeepMind to create a single unit, Google DeepMind, led by Demis Hassabis. [4] Ghahramani moved into the new organization as a Vice President of Research and a member of its research leadership team. [4] He has since been a co-author on the report describing Google's Gemini 2.5 generation of multimodal models, released in 2025, which his Google Scholar profile already lists among his most cited papers. [5] He has continued to speak publicly on behalf of Google DeepMind, including at industry events on the future of machine learning.
Ghahramani was elected a Fellow of the Royal Society (FRS) in 2015, the United Kingdom's national academy of sciences, in recognition of his work in probabilistic machine learning. [3] In 2021 the Society awarded him its Milner Award and Lecture, described as Europe's premier prize for outstanding achievement in computer science, "for his fundamental contributions to probabilistic machine learning." [3]
He has been deeply involved in the governance of his field. He served as program chair or general chair of the leading machine learning conferences, including AISTATS, ICML and NIPS (now NeurIPS), and sat on the board of the International Machine Learning Society. [4] He has served on the editorial boards of journals including the Journal of Machine Learning Research, the Journal of Artificial Intelligence Research, the Annals of Statistics, Machine Learning and Bayesian Analysis, and was Associate Editor-in-Chief of IEEE Transactions on Pattern Analysis and Machine Intelligence. [4] Beyond academia he contributed to the Royal Society's 2017 report on machine learning and led the United Kingdom government's Future of Compute Review.
As of 2026 Ghahramani continues to hold his professorship at Cambridge alongside his role as Vice President of Research at Google DeepMind. [2][3]