Predictive AI
Machine Learning Interatomic Potentials · Property Prediction · Electronic Structure Learning
Our research develops predictive AI across multiple physical scales, from electron densities and interatomic interactions to molecular and materials properties, with particular emphasis on accuracy, transferability, and reliability beyond the training data.
Generative AI
Molecular Generation · Materials Generation · Condensed-Phase Generation
We develop generative models that create three-dimensional molecular structures, crystalline and porous materials, and condensed-phase configurations, with an emphasis on controllable generation toward desired properties and physically meaningful states.
Physics- and Chemistry-Aware AI
Fragment-Constrained Generation · Environment-Aware Molecular Design
We develop AI models that incorporate physical principles and chemical constraints to improve their interpretability, generalizability, and ability to navigate chemically meaningful spaces. We aim to move AI beyond statistical pattern recognition toward models that capture underlying chemical principles.
AI Infrastructure
Datasets & Databases · Benchmarks & Metrics · Software & Web Platforms· Agentic AI
We develop open and accessible infrastructure that supports the development, evaluation, and deployment of AI models for chemistry and materials science.
Related work: PropMolFlow (3D Property Guided Molecular Generation): https://propmolflow-website.vercel.app/
COFFlow (2D COF generation): https://cofflow.vercel.app/