Cooper Lorsung

I am a PhD candidate in Mechanical Engineering at Carnegie Mellon, where I work on machine learning applied to physical systems. I am part of the Mechanical and AI Lab.

Email  /  Resume  /  Bio  /  Google Scholar  /  Twitter  /  Github

I'm broadly interested in applying computational and machine learning techniques to physical systems. In the past I have worked with Density Functional Theory, Quantum Monte Carlo, Molecular Dynamics, and numerical methods. I am currently interested in leveraging modern machine learning methods such as graph neural networks, transformers, and neural operators for molecular and fluid simulations.

LLMs are used to improve Neural Operator performance by incorporating text data.

Various pretraining strategies are compared and benchmarked for PDE systems.

Physics information is incorporated into similarity and distance functions for contrastive learning.

A novel equation embedding strategy paird with a novel transformer-based architecture allows any neural operator to become physics informed.

A neural network-based method is developed to predict ion concentration in nanochannels.

PyQMC is a python package developed to perform Quantum Monte Carlo calculations in connection wiht PySCF.

We develop a general purpose framework for using Deep Reinforcement Learning to iteratively remove vertices from meshes in Computational Fluid Dynamics.

Auglichem is a data augmentation package for molecular and chemical systems.

LUNA is an uncertainty quantification method specifically designed to model uncertainty in data-scarce regions.

We implemented implicit layers in Needle. A fully functional automatic differentiation package developed in 10-714: Deep Learnign Systems.

We developed a novel algorithm for predicting adversarial missil trajectory so a response missile can intercept its flight.

Template taken from Jon Barron's website and repository.