Research
Physics-aware learning for complex subsurface systems
My research connects high-fidelity mechanics with scientific machine learning that is fast, reliable, and interpretable.
Simulation gives us physical fidelity. Learning gives us speed and adaptability. My work asks how to keep both.
Computational geomechanics
Coupled thermo-hydro-mechanical models of fractured rock, pressurized cavities, excavation, and reservoir systems using finite elements and MOOSE.
Scientific machine learning
Physics-informed neural networks, implicit neural representations, and operator learning for geometry-aware, physics-aware prediction.
Reservoirs & porous media
Poromechanics, geothermal injection-production, fracture-controlled transport, and multiscale upscaling in porous subsurface media.
Reliable AI for simulation
PDE-grounded verification and LLM-based systems that reason about physical intent, not just syntactically valid simulation code.
Current projects
Local mechanics, reservoir response, and trustworthy AI workflows.
Geothermal reservoirs
Operator learning for fractured injection-production systems
Surrogate models for rapid evaluation of coupled geothermal scenarios, trained on high-fidelity simulation ensembles.
Rock mechanics
Physics-informed cavity mechanics across complex geometry
Stress and displacement prediction for arbitrary smooth cavities embedded in heterogeneous rock.
LLMs for science
PDE-grounded intent verification
Checking whether LLM-generated multiphysics code solves the intended physical problem rather than merely running.
