Department of Computational and Data Sciences
Department Seminar
Speaker : Dr. Bharath Ramsundar, CEO, Deep Forest Sciences, USA
Title : Towards Breakthrough Generative AI for Chemistry.
Date & Time : June 18th, 2026 (Thursday), 11:30 AM
Venue : # 102, CDS Seminar Hall
ABSTRACT:
Chemistry as a field has yet to benefit from the transformative capabilities of generative AI due to extensive hallucinations, lack of common sense, and cascading errors at longer time scales. I hypothesize that for generative AI to solve problems in chemistry, it must be buttressed with domain-specific infrastructure that takes advantage of molecular structure and properties. In particular, I have taken a multi-pronged approach in three recent papers from my group at Deep Forest Sciences. First, I have built “chemical foundation models,” domain-specific large language models (LLMs) trained on molecular structure data. In joint work with Lawrence Livermore National Labs, I have introduced ChemBERTa-3, a fully-open framework for training and evaluating chemical foundation models along with an open source model, weights, and data release. Second, to overcome LLM hallucinations, I have developed an iterative feedback method to intersperse chemically grounded validation with LLM invocations. This algorithm, DeepRetro, is a state-of-art system for chemical retrosynthesis. DeepRetro’s iterative LLM calls provide pathways and reasonable experimental instructions to synthesize potential therapeutics. I have used DeepRetro to find novel synthetic pathways for complex natural products such as Ohuamine-C and erythromycin. Third, I have introduced a framework, DeepChem-DEL, to effectively leverage chemical foundation models for modeling large DNA-encoded library (DEL) datasets. This framework enables effectively modeling the underlying chemical biology to facilitate effective AI modeling. Together, these projects provide a framework and roadmap to leverage generative AI technology to enable breakthrough chemistry.
References:
[1] DeepRetro: Retrosynthetic Pathway Discovery using Iterative LLM Reasoning
https://www.nature.com/articles/s41598-026-38821-z
[2] ChemBERTa-3: An Open Source Training Framework for Chemical Foundation Models
https://pubs.rsc.org/en/content/articlehtml/2025/dd/d5dd00348b
[3] DeepChem-DEL: An Open Source Framework for Reproducible DEL Modeling and Benchmarking,
https://chemrxiv.org/doi/full/10.26434/chemrxiv-2025-f11mk
BIOGRAPHY:
Bharath is the founder and CEO of Deep Forest Sciences, which is building an AI-powered suite for drug and materials design and discovery. Bharath received a BA and BS from UC Berkeley in EECS and Mathematics and was valedictorian of his graduating class in mathematics. He did his PhD in computer science at Stanford University where he studied the application of deep-learning to problems in drug-discovery. At Stanford, Bharath created the deepchem.io open-source project to grow the deep drug discovery open source community, co-created the moleculenet.ai benchmark suite to facilitate development of molecular algorithms, and more. Bharath’s graduate education was supported by a Hertz Fellowship, the most selective graduate fellowship in the sciences. Bharath is the lead author of “TensorFlow for Deep Learning: From Linear Regression to Reinforcement Learning”, a developer’s introduction to modern machine learning, with O’Reilly Media, and “Deep Learning for the Life Sciences”. Additionally, he authored “The DeepChem Book” in collaboration with the DeepChem team, published in 2024, and is currently working on “Differentiable Physics: Machine Learning for Physical Systems”.
Host Faculty: Dr. Phani Motamarri
ALL ARE WELCOME
