Physical laws.
Learning algorithms.
Developing neural methods to describe, simulate and control physical systems.
Scientific machine learning
Learning with the equations
Physics-informed neural networks incorporate governing equations into the learning problem. My research explores these methods for differential equations, including systems with sharp transitions and coupled dynamics.
The applications range from relativistic fluid dynamics to gravitational collapse and quantum control. My broader machine learning work also includes transformer-based particle tracking and deep learning for medical imaging.
Quantum dynamics & control
Learning how systems evolve
I study neural approaches to counterdiabatic driving: control strategies intended to suppress unwanted transitions during quantum evolution.
This work connects quantum computation and metrology, including the optimization of quantum Fisher information in time-dependent many-body systems.
Relativistic physics
Neural methods for extreme systems
I apply physics-informed learning to relativistic hydrodynamics and gravitational systems. These problems test how neural methods handle discontinuities, nonlinear dynamics and eigenvalue problems.
Recent studies have explored massless-scalar-field collapse and Kerr quasinormal modes. The calculation involving Kerr employs a two-dimensional formulation that preserves the coupled radial-angular structure.