About
I’m MD Sunbeam, a Bangladeshi American research engineer in the Human-Guided Machine Learning Group at the U.S. Army Research Laboratory. I build evaluations, demonstration datasets, and research software for models that act in games, simulators, and robotic systems.
What I care about
I am more interested in how learning fails than in a clean account of how it works. Failure exposes assumptions that success can conceal: what a model actually saw, what an evaluator allowed, and whether a result came from learning or from the machinery around it.
I am also drawn to learning with limited data and compute. The desert teaches more about water than the ocean ever will. Scarcity forces a method to reveal what it needs, what it wastes, and which ideas still hold when scale cannot cover the cracks.
This is why I care about evaluation, human demonstrations, and research infrastructure. I want experiments to tell us why a system failed, not only whether it cleared a benchmark.
How I got here
The path was not carefully planned. I chose aerospace engineering after watching Interstellar. Later, the AlphaGo documentary pulled me toward AI. My career choices have often followed the latest piece of media I loved. The decisions were impulsive, but the questions lasted.
I earned a B.S. in Aerospace Engineering from The University of Texas at Austin and an M.S. in Aerospace Engineering from Texas A&M University. At Texas A&M, I studied how human gaze and demonstrations could guide a learning agent. That work led to my current research on multimodal model evaluation, post-training, and robot learning.
Outside the work
Outside work, I am usually reading, preferably with Irani chai and dark chocolate nearby. I read both fiction and nonfiction, and I return to fantasy most often.