│ Yu (Cosmo) Du • 杜宇 ├───────────────────── │ answeror@gmail.com │ │ Cosmo is a Research Scientist at │ Meta Superintelligence Labs. │ Previously he was a Principal │ Scientist & Director at Google │ DeepMind working on Gemini │ Post-training, Thinking, and Coding. │ He was a core contributor of │ Gemini 1, 1.5, 2, and 2.5. He led │ the development of Gemini's first │ large, stable, and fast async RL │ recipe that scales across model │ ladders. He developed Gemini-0801, │ Google's first #1 on the LMSYS. │ Previously he led Bard/LaMDA │ post-training, developed the first │ internal version of Bard. Before │ that he worked on Task-oriented │ Dialog, helped Duplex scale up 10x. │ │ │ 2025- ├────────── │ Meta Superintelligence Labs │ in Menlo Park │ Research Scientist │ │ │ 2025-2025 ├────────── │ Google DeepMind in Mountain View │ Principal Scientist & Director │ │ Gemini Post-training/Thinking/Coding │ │ │ 2023-2025 ├────────── │ Google DeepMind in Mountain View │ Senior Staff Research Scientist │ │ Gemini/Bard Post-training │ │ │ 2021-2023 ├────────── │ Google Brain in Mountain View │ Staff Software Engineer │ │ Bard/LaMDA Post-training │ │ │ 2019-2021 ├────────── │ Google Brain in Mountain View │ Senior Software Engineer │ │ Task-oriented Dialog Research │ (Duplex/Dialogflow) │ │ │ 2019-2019 ├────────── │ Google Brain in Mountain View │ Software Engineer │ │ Task-oriented Dialog Research │ (Duplex) │ │ │ 2017-2018 ├────────── │ Google Ads in Shanghai │ Software Engineer │ │ │ 2011-2017 ├────────── │ Zhejiang University │ Ph.D. in Computer Science │ │