Former OpenAI Researcher Barret Zoph Rejoins Google to Lead Gemini Research
Barret Zoph, a prominent AI researcher specializing in reinforcement learning and post-training, has rejoined Google as Vice President of Research to support the development of the Gemini model. This move follows his recent departures from Thinking Machines Lab and OpenAI. Zoph's return to Google strengthens the Gemini team's capabilities in post-training and reinforcement learning, which are critical for aligning and optimizing large language models. It also highlights the intense competition for top-tier AI talent among tech giants like Google and OpenAI. Zoph previously managed the post-training team and product deployment for ChatGPT at OpenAI before co-founding Thinking Machines Lab with Mira Murati. He is rejoining Google alongside other key researchers, aiming to apply his expertise to Google's flagship Gemini models.
## BACKGROUND
Post-training in large language models (LLMs) involves techniques like fine-tuning and Reinforcement Learning from Human Feedback (RLHF) to turn raw pre-trained models into safe, instruction-following assistants. Thinking Machines Lab is an AI startup founded by former OpenAI CTO Mira Murati, which focuses on open-weights models like Inkling.