Matthew William Repasky
(Postdoctoral Research Associate)
| Email: | matthew.w.repasky@nasa.gov |
| Org Code: | |
| Address: |
NASA/GSFC Mail Code 698 Greenbelt, MD 20771 |
| Employer: | UMBC Univ. of MD, Baltimore County |
Brief Bio
My primary research focus is in the application of machine learning and AI techniques to problems in planetary geophysics. I am particularly interested in improving shape-from-shading (photoclinometry) pipelines using a combination of analytical and numerical techniques. My PhD thesis covered topics in deep learning and generative modeling, including the development of a novel method for high-dimensional posterior sampling using deep generative priors and the application of deep learning diffusion models to super-resolution of lunar topography data.
Positions/Employment
Postdoctoral Research Associate
UMBC / CRESST / NASA GSFC - Greenbelt, MD
May 2026 - Present
I apply machine learning and AI methods to problems in planetary science.
Senior Data Scientist
C3 AI - Tysons, VA
March 2025 - April 2026
I developed operations research solutions to customers in the federal space, including a network rebalancing solution for scenario analysis in the Defense Logistics Agency (DLA), production scheduling optimization solutions supporting the maritime industrial base, and a generative AI natural language search tool to query DLA structured and unstructured data using agentic capabilities.
Education
- PhD, Machine Learning, Georgia Institute of Technology, 2025
- BS, Physics, Georgia Institute of Technology, 2021
Selected Publications
Refereed
2026. "Improving Lunar Topography with Deep Learning Schrödinger Bridges.", The Planetary Science Journal, 7 (6): 139 [10.3847/psj/ae6244] [Journal Article/Letter]
2026. "Deep Graph Kernel Point Processes over Networks.", Journal of Computational and Graphical Statistics, 1-41 [10.1080/10618600.2026.2652934] [Journal Article/Letter]
2026. "Multi‐Agent Reinforcement Learning for Joint Police Patrol and Dispatch.", Naval Research Logistics (NRL), [10.1002/nav.70059] [Journal Article/Letter]
2025. "Needle in a Haystack: Information Recovery in Low Signal‐to‐Noise Piezoresponse Force Microscopy Data.", Small Methods, 10 (2): 2500318 [10.1002/smtd.202500318] [Journal Article/Letter]
2025. "Consistency Posterior Sampling for Diverse Image Synthesis.", Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 28327-28336 [Proceedings]
2024. "Stage-Regularized Neural Stein Critics For Testing Goodness-Of-Fit Of Generative Models.", ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 7255-7259 [10.1109/icassp48485.2024.10445927] [Proceedings]
2023. "Neural Stein Critics With Staged L2-Regularization.", IEEE Transactions on Information Theory, 69 (11): 7246-7275 [10.1109/tit.2023.3299258] [Journal Article/Letter]