北京雁栖湖应用数学研究院

塔赫蕾·埃夫特哈里 , 胡丕丕 , 梁鑫 , 马志婷 , 哈米德·莫菲迪 , 苏春梅 , 阿克塞尔·特恩奎斯特 , 王丽 , 熊繁升 , 杨朔 , 杨武岳

张佳

2026年04月15日 14:00 至 15:00

Shuangqing-8D01

Zoom 518 868 7656 (BIMSA)

This talk presents our end-to-end AI infrastructure for life science: a closed-loop pipeline spanning multi-scale data curation, a physics-informed architecture, efficient training, and inference acceleration. We introduce a dual-architecture where Molecular Simulators serve as digital twins capturing biophysical processes with high fidelity, while Life Foundation Models act as scientific brains extracting patterns and generating hypotheses—collaborating to solve critical challenges in protein design and cellular behavior prediction. Finally, reflecting on our journey at the AI frontier, we will discuss how researchers can rapidly evolve their mindset and skillset to navigate and lead in this era of profound technological change.

Jia Zhang is an AI Scientist at IQuest Research, focusing on geometric deep learning for quantum chemistry. He develops equivariant neural architectures that bridge computational efficiency and physical rigor for molecular simulation and first-principles computation. Previously, he served as Principal Research Manager at Microsoft Research AI for Science Lab, where he led the Science Foundation Model team in exploring unified atomistic modeling; earlier in his career, he was at Microsoft Research Asia working on machine learning for industrial systems. He has published extensively in leading venues such as NeurIPS, ICML, ICLR, and Nature Communications, advancing symmetry-preserving machine learning and its scalability in atomistic modeling.