Anisotropic Machine Learning

Project Description

Most coarse-grained simulations focus on spherical, isotropic particles, often yielding incorrect results by neglecting molecular geometry. The Cersonsky lab has developed a method using machine learning to accurately predict interactions between anisotropic particles, paving the way for shaped-particle simulation.

Interns design, document, and test a software library that generates coarse-grained “beads” from atomistic configurations, performs efficient calculation of anisotropic descriptors, and enables seamless training of machine-learned potentials.

Mentor

Rose Cersonsky

Interns

• Tejas Dahiya (Fall 2025 and Spring 2026)

Lightning Talks

• Tejas Dahiya, Fall 2025