Research
Here are some selected research interests and projects I have worked on / am working on.
Machine-learning for semi-empirical quantum-chemistry methods
Building machine-learning models for quantum-chemistry applications off of generalizable, semi-empirical QM methods can be used to improve their accuracy, generalization, and data-efficiency. Inspired by models like MOB-ML and OrbNet, we developed OrbitAll, an extension of the equivariant deep neural-network model extended to a broader set of chemical space by including spin-polarization and the ability to include implicit information about an external chemical environment to calculations.
links: OrbitAll arxiv
Machine-learning methods for analyzing observed dynamics from computational chemistry simulations
Molecular dynamics is used to study the properties of important chemical systems, such as biomolecules. The observed dynamics contain a lot of potentially useful information, but are often extremely high-dimensional and noisy. Using AI models to learn from observed dynamics could allow for insights into their structure and interactions, as well as provide a potentially useful representation that moves beyond static structure.
links: mldyn repo
Applied computational chemistry studies
I’ve worked with experimental collaborators on several projects applying theory and simulation to provide insights to experimental findings and observations.