Machine Learning-Driven Molecular Photodynamics for Long-Timescale Simulations

I offer expertise in advancing molecular photodynamics simulations beyond ultrafast regimes by integrating machine learning with nonadiabatic molecular dynamics (NAMD). My work focuses on overcoming key limitations in current excited-state simulation approaches, particularly the high computational cost of on-the-fly electronic structure calculations and the difficulty of capturing long-timescale phenomena such as rare events, weak couplings, and kinetic effects.

The approach involves developing and deploying machine learning models trained on high-level quantum chemical data to accurately predict potential energy surfaces, forces, and nonadiabatic couplings. These models are embedded within NAMD frameworks to enable simulations that retain chemical accuracy while achieving orders-of-magnitude improvements in efficiency. This enables simulations to be extended into the sub-nanosecond regime and to explore complex multidimensional reaction pathways, including conical intersections and crossing seams.

This expertise is applicable to a wide range of systems and problems, including nucleobase photostability, excited-state processes in fluorescent probes, and photochemical mechanisms in materials and biological environments. The methodology is designed to be general, transferable, and scalable, making it suitable for both fundamental research and application-driven collaborations.

I am interested in collaborations involving:

• Development of ML-enhanced excited-state simulation methods

• Photochemistry and photophysics of complex molecular systems

• Integration of quantum chemistry, dynamics, and data-driven models

• Applications in materials science, biology, and pharmaceutical chemistry

This opportunity is ideal for partners seeking advanced computational solutions to study photoinduced processes across extended timescales with high predictive power.