Team
Michael Rawson, Ph.D.
AI Applied Scientist, Amazon · Seattle, WA
Applied mathematician and AI scientist. Six-plus years building and evaluating machine learning systems at Amazon, Microsoft, Meta, and PNNL. LLM and agentic systems, foundation model post-training, reinforcement learning, adversarial robustness, and uncertainty quantification.
View portfolio →
Radu Balan, Ph.D.
Professor of Applied Mathematics, UMD · College Park, MD
Professor in the Department of Mathematics and the Applied Mathematics, Applied Statistics and Scientific Computation programme at the University of Maryland, which he currently directs. Research in harmonic analysis and its applications to engineering and computer science, particularly statistical signal processing and machine learning.
Visit university page →
Marshall Rawson
AI Software Engineer, AWS · Austin, TX
Software engineer across the full stack of the machine. Five-plus years shipping production systems at AWS, Google, Nauticus Robotics, Sylphase, AFRL, and Leidos — server kernels, GNSS receivers, subsea robot autonomy, and LLM agents for enterprise supply chain.
View portfolio →Joint Work
- Zhang, L., Wang, X., Rawson, M., Balan, R., Herskovits, E. H., Melhem, E. R., Chang, L., Wang, Z., Ernst, T. Motion Correction for Brain MRI Using Deep Learning and a Novel Hybrid Loss Function. MC-Net, a UNet trained with a two-stage multi-loss function, suppresses motion artefacts in brain MRI across 57 subjects and 20,889 slices, confirmed by three blinded clinical readers. Algorithms 17(5), 215 · May 2024 · DOI 10.3390/a17050215
- Rawson, M., Rawson, M. G. Petri Net Parallel Computing Theory and Applications. In Arai, K. (ed.), Intelligent Computing (SAI 2023), Lecture Notes in Networks and Systems vol. 711, pp. 129–146. Springer, Cham. Science and Information Conference (SAI) 2023 · Sep 2023 · DOI 10.1007/978-3-031-37717-4_9
- Rawson, M., Rawson, M. G. Petri Nets for Concurrent Programming. 2022 IEEE/ACM Fifth Annual Workshop on Emerging Parallel and Distributed Runtime Systems and Middleware (IPDRM), pp. 17–24. A generalization of Petri nets with non-deterministic transition nodes, letting concurrent systems be checked for deadlocks and optimized for maximal concurrency ahead of execution. IPDRM 2022, Dallas, TX · Nov 2022 · DOI 10.1109/IPDRM56689.2022.00008 · preprint arXiv:2208.02900
- Balan, R., Okoudjou, K. A., Rawson, M., Wang, Y., Zhang, R. Optimal ℓ1 Rank One Matrix Decomposition. In Harmonic Analysis and Applications, Rassias, M. Th. (ed.), Springer Optimization and Its Applications vol. 168, pp. 21–41. Springer · 2021 · DOI 10.1007/978-3-030-61887-2_2
- nt-petri-net — the open source reference implementation of the Petri net papers above, a graph-based concurrent middleware written in Rust. MIT licensed · github.com/MarshallRawson/nt-petri-net