Jeff Dean

Who they are

Jeff Dean is co-founder and CEO of Discovery Loop — after 27 years at Google, including co-founding Google Brain in 2011, he left in 2026 to build an AI startup organized as a public benefit corporation focused on automating scientific and engineering research.

Person

Jeff Dean founded Discovery Loop in 2026 when it was a blank page — a newly incorporated public benefit corporation, no product, no team beyond co-founders — after 27 years at Google that reshaped how the industry thinks about large-scale systems and machine learning. He started at the University of Minnesota, graduating summa cum laude with a B.S. in Computer Science and Economics, then took his Ph.D. at Washington in 1996 on optimizing compilers for modern architectures — deeply technical from the start. From there: a stint at Compaq's Western Research Laboratory doing performance monitoring and information retrieval, then a brief detour through mySimon building e-commerce content retrieval systems, before joining Google in its early days. At Google he co-founded Google Brain in 2011 — the team that put deep learning at the center of everything — and later became Chief Scientist of Google DeepMind and Google Research. He co-founded Discovery Loop with longtime collaborators Sanjay Ghemawat (his MapReduce and Bigtable co-author) and Oriol Vinyals and Quoc Le, former Google DeepMind researchers. The through-line is systems at scale: compilers, distributed infrastructure, neural networks, and now automating the scientific method itself. He posts actively on LinkedIn and the Google Research blog — papers, keynotes, reflections on AI for science — and gave a keynote at KDD 2026 and spoke at CAIS 2026, so he's still very much in the public technical conversation.

Company

Discovery Loop launched in 2026 as an independent public benefit corporation — the most recent and significant move is Dean walking away from Google's Chief Scientist role after 27 years to run it. The company's stated focus is automating machine learning, science, and engineering research: in Dean's own framing, automating the scientific method. It was co-founded alongside Sanjay Ghemawat, Oriol Vinyals, and Quoc Le — a founding team with extraordinary depth in ML research. The company is newly founded as of 2026; no product, funding, or headcount figures are public yet.

Market

Discovery Loop enters a market where AI-for-science and automated research is attracting intense interest — global VC invested a record $510 billion in H1 2026, with AI absorbing the overwhelming majority. The competitive field includes well-funded foundational AI labs (OpenAI, Anthropic, xAI) as well as Google DeepMind, which Dean just left. As a public benefit corporation, Discovery Loop is structurally distinct from for-profit labs — a deliberate bet that the mission framing will matter in recruiting and partnerships.

Network

Dean's closest collaborators at Discovery Loop are Sanjay Ghemawat — his decades-long partner on MapReduce, Bigtable, and the foundational Google infrastructure papers — along with Oriol Vinyals and Quoc Le, both former Google DeepMind researchers. Within Google he reported to Sundar Pichai and worked alongside Demis Hassabis on Gemini; Thomas Kurian (Google Cloud CEO) and Urs Hölzle are peer Google executive relationships. He is also an angel investor in Perplexity, Roboflow, Sakana AI, and World Labs, giving him a broad view across the current AI startup landscape.

Sanjay Ghemawat· Co-founder, Discovery Loop; longtime Google collaborator
Oriol Vinyals· Co-founder, Discovery Loop; former Google DeepMind researcher
Quoc Le· Co-founder, Discovery Loop; former Google researcher
Demis Hassabis· Google DeepMind colleague; Alphabet Chief Scientist and Chair of Google DeepMind
Sundar Pichai· CEO, Google and Alphabet — Dean's former reporting line

How they likely show up

27-year tenure at Google, from early engineer to Chief Scientist → thinks in decade-long arcs; unlikely to be chasing quarterly metrics at Discovery Loop.
Co-founded Google Brain in 2011 while already a senior engineer, then grew it into Google's central AI organization → builds institutions, not just products; expect him to be thinking about team structure from day one.
Founded Discovery Loop with Ghemawat, Vinyals, and Le — all deeply technical co-founders, no business-side co-founder named → the early culture will be research-led and technically demanding.
Active LinkedIn and Google Research blog author on AI research, performance optimization, distributed systems, and AI for science → comfortable being publicly visible and intellectually accountable; responds well to substantive technical engagement.
Angel investor in Perplexity, Roboflow, Sakana AI, and World Labs → has a portfolio view of the AI landscape; will be pattern-matching your work against companies he already knows.
Left Google the same day Discovery Loop launched — his own account includes a yoga session on his final Google day before midnight startup launch → high energy, not prone to slow transitions.

Conversation tips

→ Come in with a specific technical position, not a general AI enthusiasm — his public writing is precise and systems-oriented; he'll engage with concrete claims and lose interest in vague ones.
→ Ask about the public benefit corporation structure — it's a deliberate choice that distinguishes Discovery Loop from for-profit labs, and he'll have thought carefully about why.
→ Reference his KDD 2026 keynote or a specific paper he's posted about — he notices when people have actually read the work.
→ If you have a view on AI for science or automated research, say it plainly — he's been thinking about this longer than almost anyone and will push back if he disagrees.
→ Don't ask him to compare Discovery Loop to Google; ask what the new structure lets him do that the old one didn't.

Toolbox

Openers

Open on Discovery Loop's public benefit corporation structure — he and Ghemawat chose that form deliberately, and it's a pointed signal about what he thinks for-profit labs get wrong about the mission.
Reference his 2011 Google Brain founding — he built that team from scratch inside a company that wasn't yet a research org, which is directly relevant to what he's doing now at Discovery Loop.
Mention his angel bets in Sakana AI and Perplexity — two very different approaches to AI research and deployment — and ask what the through-line in his portfolio tells him about where the field is heading.

Discovery questions

Google Brain started as a skunkworks and became the center of gravity for AI at one of the world's largest companies — what did you learn about scaling a research culture that you're carrying into Discovery Loop?
You've framed Discovery Loop around automating the scientific method — where does human scientific judgment remain irreplaceable, and where do you think automation will actually surprise people?
The public benefit corporation structure is unusual for an AI lab — what constraints or freedoms does it create that a standard C-corp or nonprofit wouldn't?

Avoid

Don't treat him as a Google spokesperson or ask him to weigh in on Alphabet's current AI strategy — he left that role in 2026 and is now building something explicitly independent of it.

Make it yours