Research
I study how learned models of physical dynamics can guide control and optimization in water-energy systems and precision materials manufacturing.
Three research directions form one sequence. A shows data center water-energy coupling and a sequence of evolving material structures. B shows a neural operator mapping an input field to a predicted state. C shows an illustrative state trajectory under a constraint and an input schedule. The solid arrows connect learning, optimization, and physical action. A dashed arrow returns decision error to model training.
Physical dynamics
I investigate which physical states and time scales a model must resolve to predict responses to operating or processing inputs.
Water-energy systems
I investigate how computing demand interacts with cooling and energy supply over time in multi-energy systems serving data centers. I study how these dynamics affect operating decisions and the resulting energy and water use alongside carbon emissions.
Materials processing
I work with Professor Fengqi You on dynamic modeling and optimization for precision materials manufacturing. I study how material structure evolves under changing processing conditions and how to optimize those conditions for a target structure.
Learning dynamics
I develop neural operators to predict distributed dynamics under changing inputs and initial conditions. I study how prediction errors accumulate over a rollout and how accurately learned operators represent the sensitivities needed to optimize those inputs.
Model2Action
I study how errors in learned dynamics affect control actions and optimization outcomes, particularly when operating constraints are active. I investigate how decision regret and constraint violation can guide model training, and evaluate the resulting controllers through closed-loop cost.