GitHub LinkedIn

B.L. DuBois

ML Scientist | Mathematician | Former Green Beret

Education

BS, University of Connecticut

Class of 2026

Mathematics, GPA: 3.73.

Graduate coursework: Modern Analysis, Abstract Algebra I–II, Measure Theory, Functional Analysis.

Employment

Researcher, Department of Physics, University of Connecticut

2024 – 2026

  • Led an independent research program applying ML solutions to problems in astrophysics, with a focus on deciphering the 3D structure of the Milky Way Galaxy's Central Molecular Zone (CMZ).
  • Work culminated in creation of the IRIS project, yielding a publication and open-source code release.
  • Designed neural networks and PyTorch training pipelines involving convolutional neural networks (CNNs), generative adversarial networks (GANs), autoencoders, transformers, and neural fields.
  • Used a purpose-designed CNN to infer 3D spatial structure in radio/microwave spectral-line observations by training on a massive custom dataset of galactic simulations.
  • Developed code to synthetically observe galactic/hydrodynamic simulations differentiably in PyTorch by simulating radiation physics.

Green Beret + Other Roles, US Army

2017 – 2024

  • Among other roles, served as a Senior Special Forces Communications Sergeant (18E) on an Operational Detachment Alpha (ODA).

Publications

IRIS: Deciphering Spectral-Line Imagery of the Galactic Center by Machine-Learning on Simulations

B.L. DuBois, C. Battersby, J.C. Baade, D.R. Lipman, H.P. Hatchfield, J. Sullivan, R. Bentley, S. Reissl, R.S. Klessen, V.F. Ksoll, M.C. Sormani, Z.-X. Feng, A. Ginsburg, and R.G. Tress

arXiv preprint, arXiv:2607.01338 [astro-ph.GA], 2026. doi:10.48550/arXiv.2607.01338

3D CMZ. V. A New Orbital Model of Our Galaxy's Center, Informed by Data Across the Electromagnetic Spectrum

D.R. Lipman, C. Battersby, D. Walker, M. Clavel, B.L. DuBois, A. Ginsburg, J.D. Henshaw, R.S. Klessen, E.A.C. Mills, F. Nogueras-Lara, M.C. Sormani, and R.G. Tress

The Astrophysical Journal, vol. 1002, no. 1, p. 24, May 2026, Article 24. doi:10.3847/1538-4357/ae561f

Code Releases

Imagery Reversion Informed by Simulation (IRIS)

2026

github.com/bldubois/IRIS

  • A large ML project in astrophysics, released in tandem with the IRIS paper.
  • Uses a deep CNN to infer spatial structure in microwave spectral-line observations of the galactic center by training on galactic simulations.
  • Results yielded a new "bird's-eye" perspective of our own galaxy's Central Molecular Zone (CMZ).
  • Also includes code for the synthetic observation of spectral lines in galactic/hydrodynamic simulations by modeling radiation physics.
  • The IRIS synthetic-observation code is fully differentiable in PyTorch and, through GPU acceleration, provides speedups in excess of 10,000x against established counterparts.

Skills

Coding
  • A variety of imperative and functional languages, including Python, C, Haskell.
Machine Learning
  • PyTorch, TensorFlow, Keras; familiarity with a broad variety of training techniques and architectural heuristics including convolutional neural networks (CNNs), generative adversarial networks (GANs), autoencoders, transformers, and neural fields.
High-Performance Computing
  • Scientific and parallel/distributed computing techniques for high-performance CPU and GPU clusters (HPCs), including with NumPy, CuPy, MPI, mpi4py, Slurm.
Languages
  • Mandarin Chinese, basic familiarity; formerly 3/3 on the Oral Proficiency Interview (OPI).
Miscellaneous
  • Git version control, typesetting with LaTeX and TikZ.