Jeff Dean
Co-founder and CEO, Discovery Loop
Jeff Dean combines distributed systems engineering with large-scale machine learning research. He co-authored MapReduce and contributed to DistBelief, TensorFlow and sparse mixture-of-experts research during his long Google career. In 2026 he co-founded Discovery Loop, which aims to automate experimental loops in science and engineering, beginning with machine learning. Its investor Radical Ventures identifies him as the company’s CEO.
Get AI Secret’s free daily briefing, including Jeff Dean and the wider AI industry.
At a glance
Jeff Dean combines distributed systems engineering with large-scale machine learning research. He co-authored MapReduce and contributed to DistBelief, TensorFlow and sparse mixture-of-experts research during his long Google career. In 2026 he co-founded Discovery Loop, which aims to automate experimental loops in science and engineering, beginning with machine learning. Its investor Radical Ventures identifies him as the company’s CEO.
- Organization
- Discovery Loop
- Known for
- MapReduce; TensorFlow; scalable machine learning
- Previously
- Chief Scientist, Google Research and Google DeepMind
Sources checked . Selected career background and milestones.
Career, in context.
Selected research, leadership and career milestones, newest first.
Co-founds Discovery Loop
Dean, Sanjay Ghemawat, Quoc Le and Oriol Vinyals announce Discovery Loop, a public benefit corporation developing AI systems to automate experimental loops in science and engineering.
Source: Wilson Sonsini: Discovery Loop launch ↗Becomes Google’s Chief Scientist
Google announces Dean’s appointment as chief scientist for Google Research and Google DeepMind, responsible for strategic technical AI projects during the creation of the combined DeepMind unit.
Source: Google: Google DeepMind organization announcement ↗Co-authors sparse mixture-of-experts research
Dean joins Noam Shazeer and collaborators on a paper introducing a sparsely gated mixture-of-experts layer, increasing model capacity while activating a subset of experts per example.
Source: Shazeer et al.: Sparsely-Gated Mixture-of-Experts ↗Helps lead distributed TensorFlow release
Google releases TensorFlow 0.8 with distributed training support. The announcement credits Dean and Rajat Monga with engineering leadership, alongside contributors responsible for implementation, testing and models.
Source: Google Research: TensorFlow 0.8 ↗Co-announces TensorFlow’s open-source release
Dean and Rajat Monga introduce TensorFlow as Google’s second-generation machine learning system, releasing a standalone library and tools intended for research and production across computing environments.
Source: Google Research: TensorFlow open-source release ↗Co-authors the DistBelief paper
Dean and collaborators describe a distributed deep-learning framework and training algorithms designed for very large networks, connecting cluster computing with the practical scaling of neural models.
Source: Google Research: Large Scale Distributed Deep Networks ↗Co-authors MapReduce
Dean and Sanjay Ghemawat publish a programming model and implementation for large-scale data processing, handling parallel execution, data partitioning and failures across clusters of commodity machines.
Source: Google Research: MapReduce ↗
Show earlier historyHide earlier history
The latest on Jeff Dean
Newest firstRelated reporting
Briefs, articles and relevant sections, newest first. Explore older coverage below.
More coverage
Opus Hacked the Test
Plus: Agents Shift Already Began, Junior AI Researchers Look Replaceable
