01

AI that learns physics

I want learning-based models to pick up governing physics, geometry, and coupled processes, not just statistical patterns in data.

02

INRs, operator learning, LLMs, and PINNs

I use implicit neural representations, operator learning, LLM-based systems, and physics-informed neural networks as different ways to build physical structure directly into AI models.

03

Mechanics, poromechanics, and bio-inspired systems

Most of these ideas are grounded in computational mechanics, poromechanics, multiscale transport in porous subsurface media, and bio-inspired mechanics, often alongside MOOSE-based simulation workflows.

What connects these projects is a simple question: how can we get learning-based models to respect physics instead of just fitting data? I care most about problems where geometry, constitutive behavior, and coupled processes really matter, and where better physical structure can make AI tools more reliable for mechanics and Earth science.