PES-Learn

PES-Learn is an open-source Python package that generates machine learning models for molecular potential energy surfaces. In 2025 I published a paper in the journal Molecules that outlines some features that I added to the program. These features include a new machine leanring model type (kernel ridge regression), new data generation methods, benchmarking, and the ability to compute gradients.

Working on this project cultivated my skills as a Python developer and machine learning engineer. In addition to learning about a variety of machine learning methodologies, I matured my knowledge of Python packages including data science packages such as NumPy and Pandas, and machine learning packages like PyTorch and scikit-learn. The primary challenge that I faced working on this project was manipulation of molecular representations. The problem at hand is finding a way to accurately represent a molecular geometry so it can be manipulated through a machine learning model. There are a variety of ways to represent a geometry, transforming them in pre- and post-processing for user experience was a fun challenge to overcome.

The aformentioned publication can be found here and the GitHub repository for PES-Learn can be found here.