Tatiana U.

I have over 4 years of experience in Machine Learning and Software Development and 7 year of research across startups R&D and academia, specializing on solving hard underdefined problems in AI for scientific applications and physical world: medical, biological, and physics grounded data. I excel in areas where there is no existing roadmap and figuring out 0 → 1 is the central problem. Most recently, I've been leading ML perception R&D as a Founding Engineer and leading ML algorithm developer at a 5 person pre-seed startup. There I developed a core perception system from idea to a working system ready for investor demo and customer testing. I designed a physics-based algorithm that achieved 90%+ classification accuracy in cluttered environment with moving targets, on a timeline and dataset size that ruled out standard approaches. I owned the full technical scope: problem framing based on implementation complexity and customer constraints, data collection strategy, feature engineering, physics interpretation for determining stable features, core architecture, stability verification, and performance analysis through mathematical modeling and experiments. I conducted research on integrable models of quantum field theory in A. Belavin's group (one of the founders of conformal field theory, a field related to string theory) in Moscow, and did my PhD in astrophysics at Johns Hopkins University studying atmospheric escape. I transitioned to industry and currently work in ML/AI. I am interested in working on hard problems.