名师讲座

报告题目：Toward a Researcher-Centered AI Assistant: Knowledge-Driven Agentic Systems for Computational Chemistry

报告时间：2026-08-08 10:15

报告人： Tao Jiang

Institut de Chimie de Nice, Université Côte d'Azur & 3iA Côte d'Azur

报告地点：卢嘉锡楼202报告厅

转播地点：翔安校区能源材料大楼3号会议室1；漳州校区生化主楼307教室

报告摘要：

Large language models have opened up new possibilities for accelerating scientific research, and researchers have been working to build AI assistants for scientific research. Yet the capabilities of these models are evolving so rapidly that building useful systems remains a moving target: each system must be designed around LLM’s capabilities at a given time. The systems presented in this talk reflect exactly such a trajectory.

We have explored a wide spectrum of architectural designs in response to shifting model capabilities. Early efforts focused on grounding model outputs in reliable, curated scientific corpora1. As tool-use became dependable, we built more elaborate frameworks—with planners, executors, and critics—to let agents discover and refine their own workflows2. More recently, as foundation models grew stronger, we experimented with a deliberately lighter architecture: minimal scaffolding, but with heavy emphasis on capturing the researcher's domain knowledge and preserving project-level memory across runs.

Finally, I will show our latest effort on integrating these directions into a single web-based interface that unifies literature access, agentic execution, cluster job management and persistent memory.

1. L. Pradi, T. Jiang, M. Feraud, M. Bekbergenova, Y. Taghzouti, L.F. Nothias, “An AI Pipeline for Scientific Literacy and Discovery: a Demonstration of Perspicacit´e-AI Integration with Knowledge Graphs”, ISWC-C 2025.

2. M. Legrand, T. Jiang, M. Feraud, B. Navet, Y. Taghzouti, F. Gandon, E. Dumont, L.F. Nothias, “Mimosa Framework: Toward Evolving Multi-Agent Systems for Scientific Research”, arXiv preprint arXiv:2603.28986, 2026

报告人简介：

Dr. Tao Jiang is a CNRS Research Engineer at Institut de Chimie de Nice, France. He received his B.E. and M.E. in Chemical Engineering from Dalian University of Technology and Dalian Institute of Chemical Physics, and obtained his Ph.D. in Physics from Technical University of Denmark in 2010. He completed successive postdoctoral appointments at Ecole Normale Supérieure de Lyon and Utrecht University. His research spans computational catalysis, multi-scale QM/MM molecular dynamics, biomolecular simulation and protein crystallization prediction. His recent work centers on AI for computational chemistry, developing autonomous multi-agent frameworks including Mimosa, Magnolia and Perspicacité for metabolomics mass spectrometry analysis, enzyme activity modeling and intelligent literature mining. Skilled in multiple programming languages and HPC administration, his research outputs have been published in top international journals, and he has delivered numerous oral and poster presentations at worldwide academic conferences.

欢迎老师同学们积极参加!

化学化工学院

表界面化学全国重点实验室