Pre-Print
Toward Goal-Agnostic Joint-Embedding Predictive Control of Partial Differential Equations
// Research
Research directions & publications
My work spans scientific computing, dynamical systems and deep learning, with an emphasis on developing models that are provably reliable and practically deployable for performance optimization and control, specifically on resource constrained hardware. Learn more about my current interests below!
neural-operators
Neural Operators for multiphysics problems
Learning infinite-dimensional operators that map initial conditions to time dependent trajectories for multiphysics systems, enabling for cheap-to-inference predictions of future dynamics.
FNODeepONetOperator Learning
JEPA
JEPA world models for PDE Control
Im actively exploring how JEPA world models can lend themselves to goal agnostic, flexible and robust controllers for PDE systems.
JEPAWorld Modeling
Neural-MPC
Neural Surrogate Model Predictive Control
Replacing expensive first-principles solvers inside MPC loops with structure-preserving neural surrogates, targeting real-time feasibility for systems governed by stiff or high-dimensional dynamics.
MPCSurrogate ModellingReal-Time Control
All Publications