Recorded video link: https://vshare.sjtu.edu.cn/open/d22522284388cd356381024fab757a1670da792ac5b02d79f811c49112e32e88

Abstract:

AI for Science (AI4S) is undergoing widespread adoption, driven by transformative changes in the foundational infrastructure supporting reading, computing, and experimentation. Fostering this shift towards a platform-driven research paradigm requires an open-source, community-driven approach to build AI4S practice. I will introduce my exploration process and practice in this direction over the past 10 years, and present this model in action: 1. Literature Mining & Annotation: Accelerating knowledge synthesis; 2. Model, Software & Agent Co-Creation: Powering collaborative intelligence; 3. Lab Automation & Smart Infrastructure: Enabling AI-driven experimentation.

Bio:

Linfeng Zhang is the founder and Chief Scientist of DP Technology, as well as the Director of the AI for Science Institute, Beijing. He holds a Bachelor of Science degree from Peking University and a Ph.D. in Applied Mathematics from Princeton University. Linfeng has long been dedicated to interdisciplinary research in AI for Science, achieving significant results in machine learning, computational physical chemistry, materials, and drug design. He has been consistently listed in Stanford University's "World's Top 2% Scientists" ranking for multiple years. As a core developer, Linfeng has led the creation of a series of micro-scale simulation algorithms and corresponding open-source software. In 2020, DeePMD won the ACM Gordon Bell Prize, the highest honor in high-performance computing, and was also selected as one of China's Top 10 Scientific Breakthroughs of 2020 by academicians of the Chinese Academy of Sciences and Engineering. In early 2025, Linfeng and his team released Uni-3DAR, the world's first cross-scale 3D large model, unifying microscopic and macroscopic 3D worlds through autoregressive modeling.