Research

Discontinuous, hybrid and stabilized finite element formulations for flow in heterogeneous media.

I investigate multiscale problems, hydromechanical coupling and double-porosity and double-permeability models, with applications in rocks and reservoirs.

Phase equilibria and thermophysical properties, using equations of state, activity-coefficient models and fitting to experimental data.

Applications include PVT properties of petroleum fluids, CO₂ mixtures and biofuel flash points. In the context of the energy transition, thermodynamic modeling also applies to carbon capture, utilization and storage (CCUS) and underground hydrogen storage (UHS).

I study how to incorporate physical information into machine learning and integrate learned models into scientific simulations.

On-demand learning

Reusing equilibrium states and sensitivities to accelerate calculations during simulation. New full evaluations are performed when an approximation fails the acceptance criterion.

Reactive transport study with ODML

Hybrid thermodynamics

Neural models with material balance preserved by construction and training guided by thermodynamic relations such as Gibbs–Duhem. Applications include flash calculations and flash points.

GibbsFlashNN study

Numerical methods and physical information

Coupling learned models with solvers, including neural constitutive relations and automatic differentiation. I assess accuracy, physical residuals and computational cost against reference solutions.

Differentiable thermodynamics tools

Coupling flow, species transport and geochemical reactions in heterogeneous porous media.

Studies integrate Reaktoro and Firedrake to track fluid composition and mineral dissolution and precipitation.

Bayesian calibration, sensitivity analysis and uncertainty propagation in computational models.

I study parameter identification from observations and its effects on predictions, with applications in dynamical systems and epidemiological models.