Geometric deep learning · Scientific machine learning · Neural surrogates · Cardiovascular modelling

Geometric deep learning
for science and health.

I am a PhD candidate at the University of Twente, working on geometric deep learning for science and health. I develop neural surrogates for complex physical simulations, with a focus on cardiovascular blood flow, generalization, and learning from limited data. Explore our neural surrogate works in the interactive viewers for blood-flow velocity and wall shear stress.

My research is part of the EU Horizon project VASCUL-AID, supervised by Dr. Jelmer Wolterink, Prof. Dr. Christoph Brune, and Prof. Dr. Kak Khee Yeung. I collaborate with Amsterdam UMC and TU Delft, and am a visiting PhD researcher at CIMIM, University of Manchester, with Prof. Dr. Alejandro Frangi.

Previously, I worked on medical imaging at Hemolens Diagnostics under Dr. Tomasz Konopczyński and was a visiting student at NVIDIA. I earned my BSc in Algorithmic Computer Science and MSc in Artificial Intelligence at Wrocław University of Science and Technology, with both theses supervised by Prof. Dr. Maciej Zięba.

I am a member of the ML in PL Association and a co-organizer of the London Geometry and Machine Learning Summer School (LOGML). Outside research, I enjoy climbing, skiing, hiking, and photography.

Figure 1: CFD reference and model predictions under vessel rotation, comparing equivariant AB-GATr with non-equivariant AB-UPT

Figure 1: Rygiel et al., 2026 · CC BY-NC-SA 4.0

Featured research / AB-GATr

2026

When does equivariance help?

Learning fluid dynamics from geometry. We study how orientation and symmetry shape neural surrogate performance, and introduce a scalable geometric algebra transformer for coupled surface and volume predictions.

Research interests

01

Geometry & symmetry

Building inductive biases into learning on meshes, point clouds, and geometric structures.

02

Neural physics surrogates

Making complex simulations efficient, scalable, and useful when training data is limited.

03

Cardiovascular imaging

Connecting medical images, vascular geometry, and patient-specific blood-flow modelling.

A few more things I work on

Selected publications

All publications

Recently

Research grants

  • Young Talent Research Grant
    Aneurysm-NL · AI-supported in-silico observational study of aneurysmal hemodynamics.
  • Snellius computing grant · EINF-16639
    Dutch Research Council (NWO) · Development of data-efficient geometrical neural surrogates for complex physics simulations.
  • Snellius computing grant · EINF-10199
    Dutch Research Council (NWO) · Longitudinal Study of Hemodynamic in Abdominal Aortic Aneurysms.