Barret Zoph

Barret Zoph joined Google Brain in 2016 and made a name automating the design of neural networks. "Neural Architecture Search with Reinforcement Learning," posted with Quoc Le in November 2016, trained a recurrent network to propose network designs and rewarded it by how accurate the resulting models were; the designs it found matched the best hand-built image and language models. In January 2021 Zoph co-wrote the Switch Transformers paper with William Fedus and Noam Shazeer, which routed each token to one expert in a sparse mixture-of-experts model and scaled it past a trillion parameters.

Zoph moved to OpenAI in 2022 and led post-training, the stage that turns a pretrained model into a chat assistant; the o1 system card lists Zoph among its executive leadership. On 25 September 2024, the day Mira Murati resigned, Sam Altman announced that Zoph, then a vice president of research, and chief research officer Bob McGrew were leaving too. Zoph called it a "natural point for me to explore new opportunities outside of OpenAI." In February 2025 Zoph became chief technology officer of Murati's Thinking Machines Lab.

The exit from Thinking Machines is disputed. On 14 January 2026 Murati posted "We have parted ways with Barret Zoph" and named Soumith Chintala CTO. Within the hour OpenAI's Fidji Simo announced that Zoph, co-founder Luke Metz and Sam Schoenholz were rejoining OpenAI, a move "in the works for several weeks." Murati told staff that Zoph had been fired for "unethical conduct," and sources close to Thinking Machines told Wired that Zoph had shared confidential information with rivals. Simo told OpenAI staff that the company did not share those concerns. Zoph's account, reported by the Wall Street Journal, was that the dismissal came only after Murati learned of Zoph's plan to leave. Back at OpenAI, Zoph led enterprise sales for about five months and left in June 2026. In August 2026 Google said Zoph would join as a vice president of research, bringing "RL and post-training expertise to Gemini."