Hi, I'm Marshall!
Latest Writing:
With my Math PhD recently finished, I am looking for research opportunities in interpretability of AI systems. In particular, I am looking to understand AI mechanistically with the goal of uncovering the algorithms learned by models like ChatGPT, Claude, or Gemini.
At Tufts University I was advised by James M. Murphy and Abiy Tasissa. My dissertation is entitled Local Sparse Representations: Connections With the Delaunay Triangulation and Dictionary Learning in Wasserstein Space.
I like to read, write, take photos, watch movies, and practice Kyudo.
In the past I attended the University of Wisconsin-Madison with a double major in math and computer sciences (yes sciences). I was fortunate to have advising from Sigurd Angenent and Dave Anderson.
Contact: marshallm (at) protonmail (dot) ch
Website design inspired by the delightfully simple aesthetic of Shoji.