I use techniques from mechanistic interpretability and (at least occasionally) topological data analysis to better understand AI models and large unstructured datasets. I’m currently at BluelightAI, working on tools to better monitor and evaluate AI systems. If you’re interested in collaborating or just talking about any of these ideas, reach out! (Contact info is below.)

I graduated from the University of Pennsylvania in May 2020 with a PhD in Applied Mathematics and Computational Science, under the direction of Robert Ghrist. My dissertation work introduced and developed the theory of sheaf Laplacians. These are discrete Laplacian operators associated with a sheaf on a cell complex (often just a graph), and have been used in a number of different ways, including for network dynamics, graph neural networks, deep space networking, and feature selection.

Before joining BluelightAI, I was a Visiting Assistant Professor in the Mathematics department at Ohio State university, as part of the Topological and Geometric Data Analysis group.