Projects

We develop a unified computational framework that integrates quantum mechanical calculations (DFT/TDDFT), classical molecular dynamics, and continuum models. The framework enables the study of emergent phenomena in complex materials that span multiple length and time scales, including thermal transport, mechanical behavior, and electronic properties of nanostructured systems.

This project develops physics-informed neural network potentials (NNPs) and graph neural network (GNN) based force fields that achieve near-DFT accuracy at a fraction of the computational cost. We apply active learning strategies to systematically improve transferability across composition and configuration spaces, enabling high-throughput screening of novel functional materials including solid electrolytes and catalysts.

Understanding protein folding, conformational dynamics, and molecular recognition requires sampling rare events on biologically relevant timescales. We develop and apply advanced sampling techniques — including metadynamics, replica exchange, and machine learning-guided collective variables — to study protein-ligand binding, intrinsically disordered proteins, and membrane protein conformational changes.