Garrett Mulcahy

I am a sixth year PhD candidate in the Mathematics Department at the University of Washington, Seattle, where I have the good fortune of being advised by Prof. Soumik Pal. I expect to graduate in June 2027.

I am on the academic job market for positions that start in Fall 2027.

My main research interest is the mathematical foundations of machine learning models, particularly flow-based models. To this end, I use tools from optimal transport and its entropic regularized counterpart, the Schrödinger bridge, in addition to tools from stochastic calculus and probability theory.

My recent projects have involved developing approximations to the entropic regularized optimal transport plans (Schrödinger bridges) in the small and large temperature regimes and approximating gradient flows in the Wasserstein space.

I did my undergraduate studies in Mathematics and Statistics at Purdue University, where I was generously mentored by Prof. Thomas Sinclair.

Publications and Preprints

• Mulcahy, G. (2026) Noising-Denoising by Large Temperature Schrödinger Bridges. Arxiv (Preprint)
• Mulcahy, G., Pal, S. (2025) Diffusions Approximations to Schrödinger Bridges on Manifolds. Arxiv (Preprint, submitted)
• Agarwal, M., Harchaoui, Z., Mulcahy, G., and Pal, S. (2026). Langevin diffusion approximation to same marginal Schrödinger bridge. Journal of Functional Analysis, 291(1):Paper No. 111494, 27. Journal, Arxiv
• Mulcahy, G. & Sinclair T., Malnormal matrices, Proc. Amer. Math. Soc. 150 (2022), no. 7, 2969-2982
• Mulcahy, G., Atwood, B., & Kuznetsov, A. (2020). Basal ganglia role in learning rewarded actions and executing previously learned choices: Healthy and diseased states. PLOS ONE 15(2): e0228081. https://doi.org/10.1371/journal.pone.0228081

Conference Paper and Technical Reports

• M. Landajuela, C. Shing Lee, J. Yang, R. Glatt, C. Santiago, T. N. Mundhenk, I. Aravena, G. Mulcahy, B. K. Petersen, A unified framework for deep symbolic regression. Thirty-sixth Conference on Neural Information Processing Systems, 2022, NeurIPS 2022.
• Mulcahy, G., Ehrhart, B, Towards Better Modeling the Probability of Hydrogen Ignition, Sandia National Laboratories, Albuquerque, New Mexico, 2025, SAND2025-13516.
• Mulcahy, G., Brooks, D.M., Ehrhart, B, Using Bayesian methodology to estimate liquefied natural gas leak frequencies, Sandia National Laboratories, Albuquerque, New Mexico, 2021, SAND2021-4905.

Contact Information

My email is gmulcahy @ uw . edu.