Harsh Mehta
Harsh Mehta is an AI researcher and engineer based in the San Francisco Bay Area who has worked on large language model development at leading AI organizations. He served as a Member of Technical Staff at Anthropic from December 2024 to December 2025, focusing on AI R&D, and previously spent approximately ten years at Google and Google DeepMind, where he co-led the Gemini project and contributed to foundational work on transformers, state space models, and reinforcement learning. He is known for publications on convexifying transformers, long-range language modeling via gated state spaces, and the BIG-bench benchmark, and has recently departed Anthropic to start a new venture.
“"The jump from Sonnet 3.5, 4.0, 4.5 has been materially better in terms of like where does it get stuck? Where does it need oversight? Even the tiniest reduction in oversight can lead to like large amounts of token spent and just effective outcomes being generated."”
Source→“"Harsh was kind of initiated at the automated pre-training project in Anthropic... Dario had this essay, Machines of Loving Grace, which was very inspiring."”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.