Anthropic Hires UC Berkeley Department Chair, Snags Four Top Talents in Two Weeks

On July 1, Jelani Nelson, chair of the EECS Computer Science Division at UC Berkeley, announced he is joining Anthropic on academic leave while retaining his faculty position. Nelson is a theoretical computer scientist specializing in streaming algorithms, dimensionality reduction, and randomized algorithms, whose research is crucial for improving large model training efficiency and data compression. An Anthropic spokesperson confirmed he will join the pre-training team, focusing on Claude's core capabilities.

Background: Anthropic's Talent Offensive

Nelson's move is part of Anthropic's recent aggressive hiring spree. On June 18, Noam Shazeer, a co-author of the Transformer paper and co-lead of Gemini, left Google to join OpenAI. On June 19, John Jumper, 2024 Nobel Prize in Chemistry winner and AlphaFold developer, announced he was leaving DeepMind for Anthropic (likely starting next year due to non-compete clauses). On June 24, Gemini core researchers Jonas Adler and Alexander Pritzel were also reported to be following suit. In just two weeks, Anthropic has attracted at least four top talents: a Nobel laureate, two Gemini core researchers, and a department chair.

Key Details: Jelani Nelson's Academic and Industry Background

• Academic Record: Nelson earned dual bachelor's degrees in Computer Science and Mathematics from MIT, and a PhD in Computer Science from MIT in 2011, with his dissertation "Sketching and Streaming High-Dimensional Vectors" winning the MIT Outstanding Thesis Award. He held postdoctoral positions at Berkeley and Princeton, joined Harvard in 2013, moved to UC Berkeley in 2019, became Computer Science Division Director in July 2024, and was promoted to Department Chair in July 2025.
• Research Focus: Streaming algorithms, dimensionality reduction, and randomized algorithms, centered on "processing the largest data with the least memory and computation." His team proved the memory lower bound for approximate counting problems and co-proved the optimality of the Johnson-Lindenstrauss lemma, a theoretical foundation for vector retrieval and embedding compression.
• Industry Experience: From 2021 to June 2025, Nelson also worked as a research scientist at Google, making this move another instance of talent drain from Google.
• Social Contribution: In 2011, he founded AddisCoder, a free programming summer camp in Ethiopia, which has trained nearly 700 students, earning him the ACM Lawler Humanitarian Contributions Award.

Reactions and Impact

• Academic and Industry Response: YC President and CEO Garry Tan expressed amazement at Anthropic's appeal on social media. Nelson's public lecture video (Harvard YouTube channel "Advanced Algorithms") has over 21 million views, drawing widespread attention to his move.
• Impact on Google: Within June, Google lost core talents like Noam Shazeer and John Jumper, causing Alphabet's stock to drop and investors to question its talent retention capabilities.
• Trend: Nelson's "academic leave" arrangement, retaining his faculty position, reflects a growing "half-in-industry" model (e.g., Fei-Fei Li's leave to join Google Cloud in 2017). For AI companies, this reduces hiring friction and may attract talent from his academic network.

Impact: Theoretical Talent Becomes New AI Competition Focus

Nelson's addition marks an extension of AI talent wars from engineering and product to theoretical computer science. As model sizes approach computational limits, improving algorithmic efficiency (e.g., data compression, computational complexity optimization) becomes critical. Anthropic aims to strengthen its theoretical foundations, while universities face pressure from the "brain drain" of top scholars to industry.