Research
I am in the first year of my PhD, so this page is short by design. It lists what I am actually working on rather than what I hope to work on. It will grow.
Research Interests
I am interested in machine learning that is constrained by the structure of the system it models, and in what that buys you when data is scarce, expensive, or physically bounded. Concretely:
- Physics-informed machine learning: embedding governing equations and conservation laws into models rather than asking them to infer physics from data alone.
- Agentic traffic light solutions: treating signal control as a system of interacting agents rather than a fixed schedule.
- Quantum mechanics: quantum systems both as a domain for physics-informed methods and as a source of problems that classical models handle badly.
- Feedforward neural networks: the behavior and limits of the basic architecture that most of the above is built on.
Projects
AI City Solutions for Traffic Light Coordination
I worked as part of a team on AI city solutions for traffic light coordination, submitted to the NeurIPS 2026 competition track. The problem is coordinating signals across a network rather than optimizing each intersection on its own, so that timing decisions at one light account for what they push onto the next.
This is the work that pulled me toward agentic control of physical infrastructure, and toward the question my dissertation now starts from: how far learned forward operators hold up over long horizons.
DOE Genesis Mission: Water Forecasting for Energy Planning
I work with Dr. Aniruddha Bora on Drift-Aware Initialization of Coupled Earth System Models for Scalable Subseasonal-to-Seasonal Prediction Using Agentic AI, part of the Department of Energy’s Genesis Mission under a sub-award from Pacific Northwest National Laboratory.
The goal is long-range water availability forecasts that energy providers can plan against. Coupled Earth-system models drift. Error in the initial state grows as the model integrates forward, and at subseasonal-to-seasonal ranges that growth dominates the forecast. The project uses an agentic AI framework to characterize how initial error propagates and to recommend corrections before the forecast is run, supporting hydropower planning, reservoir management, and drought preparation. The work is built on E3SM, the Department of Energy’s Energy Exascale Earth System Model.
Dissertation (in progress)
Forward Operator Use on Long-Horizon Tasks / Quantum Compute. Early-stage work on where learned forward operators break down as prediction horizons grow, and how that changes for quantum systems. It shares its central question with the Genesis Mission work above, error growth over long horizons, approached from the operator side rather than the initialization side. Advised by Dr. Aniruddha Bora.
Publications
None yet. I have not published as a first-year student. When that changes, papers and preprints will be listed here with links to the code and data behind them.