Research

Scientific computing for long-range interactions, soft matter, and machine learning

I develop mathematical models and scalable computational methods for soft and active materials. My work connects numerical analysis with physically specific models of long-range interactions, collective dynamics, and scientific machine learning.

Scalable algorithms for long-range interactions

Electrostatic and hydrodynamic calculations become more difficult in confined, partly periodic, and heterogeneous systems. We develop Ewald-type decompositions, random-batch methods, boundary-integral formulations, and hybrid solvers for these settings.

We study parameter-dependent error, computational cost, and the treatment of boundary and interface effects.

Selected publications

Current questions

Error control under stronger dielectric contrast, scalable solvers for confined geometries, and links between electrostatic and hydrodynamic algorithms.

Physical mechanisms in soft and active matter systems

We model how dielectric contrast, confinement, activity, and memory shape attraction, packing, broken symmetry, and collective motion.

We test numerical observations against electrostatic theory and reduced physical models, and specify the regime for each explanation.

Selected publications

Current questions

Contact interactions between asymmetric polarizable particles, memory in chemotactic motion, and the relation between confinement, cohesion, and helicity.

Machine learning for PDEs and materials science

We study operator learning, physics-informed methods, and reduced-order models for complex PDE systems. Current work covers pattern dynamics, nonlinear fluid evolution, parametrized PDEs, and nearly touching plasmonic structures. Machine-learning methods for materials science form part of this research direction.

Selected publications

Current questions

Generalization outside the training regime, error growth over time, and comparisons with standard numerical baselines.