MILES Seminar - Dr. Chenru Duan

Dr. Chenru Duan obtained his Ph.D. in Chemistry and Chemical Engineering models into high, roughiput computation for accelerate chemicaliscovery. he then worked as a research scientist at Microsoft, developing machine learning and computational chemistry solutions for industry. Dr. Duan is also a core organizer of the Al4Science workshop series at ICML, NeurIPS, and ACS. In 2024, he co-founded Deep Principle, a startup dedicated to realizing systematic chemical design with AI. 

Generative models are revolutionizing chemical discovery. They learn rich, symmetry-aware representations of matter and sample novel structures from syme try pleting diffies a mework that learns the ph sist butioned, reactants, transition states, and products. (ii) language-model-guided evolutionary search that combines the chemical intuition of large LLMs with genetic optimization. Together these studies demonstrate a unitying vision for generative chemistry: physics-grounded diffusion models can map reaction landscapes with DFT-level fidelity, while knowledge-rich language models can invent and prioritise candidate molecules-all through data-efficient, interpretable generation pipelines that invite human guidance via plain language.