Andrzej Dulny
PhD student in machine learning at the Chair of Data Science, University of Würzburg. I work on machine learning methods for understanding physical systems — weather, waves, and fluids.
I am a PhD student at the Chair of Data Science, University of Würzburg. My research focuses on leveraging machine learning to predict the evolution of physical systems, extract the governing physical equations from data, and enhance numerical simulations with learned components.
I work with physics-informed neural networks, neural ODEs, graph neural networks, and transformer-based models for spatial data — applied to weather prediction, climate models, wave simulations, and fluid dynamics. I also contribute to the MAGNET4cardiac7T project on simulating electromagnetic fields in 7T MRI scanners.
Selected work
-
TaylorNet PMLR · 2025learning partial differential equations directly from non-grid data
-
GrINd ECML-PKDD · 2024forecasting physical systems from scattered, non-grid observations
-
DynaBench ECML-PKDD · 2023a benchmark for learning dynamical systems from low-resolution data
-
NeuralPDE KI · 2022combining convolutions with differentiable ODE solvers to model dynamical systems