GitHub LinkedIn

B.L. DuBois

ML research Math focus Defense perspective

I'm a machine-learning scientist and mathematician. I care about machine-learning applications at the frontiers of science and technology. In the research that I've led, I've designed differentiable radiation simulations and neural image transformations for radio astronomy. See my research page for more on how this work is helping decipher the structure of the galactic center. While my expertise is technical first, my outlook is grounded in a defense background. As my research evolves, that background manifests as an aspiration to seek out problems with real-world impacts.

Research CV
Headshot of B.L. DuBois
Synthetic radio astronomy image transformations

Research

In my recent work with the Milky Way Laboratory at the University of Connecticut, I developed IRIS: an ML system for deciphering the structure of the galactic center. IRIS uses a deep convolutional neural network to transform spectral-line imagery of our galaxy observed in our edge-on perspective into a top-down perspective, and trains on synthetic data generated from galactic simulations. IRIS also incorporates a GPU-accelerated and fully differentiable "synthetic observation" code, which enables generation of the IRIS training data by simulating the observational process at up to 10,000 times the speed of comparable, CPU-based codes. Read the summary on my research page, the preprint on arXiv, or visit the code repository at GitHub.

Mathematics

I have a BS in Mathematics from the University of Connecticut. In completing my degree, I undertook substantial graduate coursework, including multiple semesters of abstract algebra and courses in modern analysis, measure theory, and functional analysis. I see mathematics as both the language in which I conceptualize ML science and a style of thought with which I approach problems in general. Beyond just building code in PyTorch, my experience leads me to believe that strong ML research and development begins with an understanding of the why and an ability to reason abstractly about the how. Mathematics is not only the language in which I understand that why but the style of thought with which I approach that how.

Perspective

Before pursuing a lifelong passion for STEM to a new career in ML science, I spent most of my 20s in the military, primarily as a Green Beret. It was, in fact, my experiences in the defense community that spurred my interest in ML research. While I began coding as a kid, I was inspired to begin self-educating in ML development in around 2019, as I became convinced of the centrality of AI to the unfolding era of great-power competition. In the intervening years, I have only grown more certain of this thesis. And even with my military career behind me, my defense experiences continue to inform the way I approach the development of AI technologies as first and foremost an issue of not only national security but human security.