Jeff Dean, Google‘s chief scientist and one of the most influential figures in the company’s AI infrastructure, is leaving after nearly 27 years to co-found a new startup called Discovery Loop. The departure was announced Wednesday as part of a broader reshuffle of Google’s AI leadership.

Discovery Loop aims to automate the experimental loops of scientific research itself, starting with machine-learning research and engineering before expanding into areas such as hardware design, drug discovery, and clean energy. The company describes itself as an independent public benefit corporation.

Dean is not leaving alone. Three other senior Google AI researchers are co-founding the venture with him: Sanjay Ghemawat, a Google Senior Fellow who worked alongside Dean on much of the company’s foundational computing infrastructure; Oriol Vinyals, vice president of research at Google DeepMind and a technical lead on Gemini; and Quoc Le, a co-founder of Google Brain and the scientist behind the AutoML-Zero project.

Despite the exits, Google is backing the new company rather than treating it as a rival. The tech giant will serve as Discovery Loop’s founding investor and cloud partner, and has agreed to supply compute power for the startup’s first year of operations. The two companies also plan to collaborate on a research framework for machine-learning systems and infrastructure.

Radical Ventures and Khosla Ventures are co-leading Discovery Loop’s seed round. The round has not yet closed, and the startup has not disclosed its valuation.

Dean joined Google in 1999 and became one of the architects behind systems such as MapReduce, Bigtable, and TensorFlow, work that helped shape how the company, and much of the tech industry, builds large-scale computing infrastructure. He later played a central role in Google’s AI push as head of Google Brain and, more recently, as chief scientist overseeing the company’s broader AI strategy.

His departure adds to a string of senior AI researchers leaving major labs in 2026 to start independent companies, a trend that has accelerated as venture capital continues to pour into AI infrastructure and research automation startups.