DAI Damai (代达劢)
Deep Learning Researcher, DeepSeek
What I most want to address is the multiple layers of disconnection that have long existed in the design industry. The first layer is the disconnection between architecture and interior. Ten years ago, I already judged that China's big...
DAI Damai (代达劢) is a Deep Learning Researcher at DeepSeek, where he serves as the infrastructure lead responsible for inference system engineering optimization and large-scale deployment. He earned his Ph.D. in 2024 from the School of Computer Science at Peking University, under the supervision of Professor Sui Zhifang at the MOE Key Lab of Computational Linguistics.
Dai has published over 20 top conference papers with more than 28,000 Google Scholar citations. In 2023, he won the EMNLP Best Long Paper Award as the third core author for 'Label Words are Anchors' - the first time a mainland Chinese institution won this award.
He is the first author of the DeepSeekMoE paper (ACL 2024), which introduced the Mixture-of-Experts architecture that underpins DeepSeek's cost-efficient model performance. He has been involved in DeepSeek's development from V1 through V3.