俞栋Yu Dong

Zhejiang University Electrical Engineering BS, Chinese Academy of Sciences Institute of Automation MS (Pattern Recognition), Indiana University Bloomington CS MS, University of Idaho CS PhD. Was among the first to apply deep learning to speech recognition at Microsoft Research, where he worked for 16+ years and became a Principal Researcher. His work on DNN-based speech recognition (2011, with Hinton, Deng Li et al.) was a landmark result that started the deep learning revolution in speech. Joined Tencent AI Lab as VP in 2017. Led research in speech processing, multimodal interaction, and LLM-based NLP. Co-authored 200+ papers and several books on speech recognition. Left Tencent at end of January 2026. His exit was widely reported as the biggest trigger for the dissolution of Tencent AI Lab: on 19 March 2026 the lab was formally disbanded and folded into the Hunyuan large-model organization, with his speech team absorbed into Bo Liefeng's multimodal division. He is reported to have returned to Seattle after leaving, with no new venture announced as of mid-2026.