José Marín Fariña

Wall shear stress and pressure distributions in the thoracic aorta are relevant biomarkers for vascular remodeling, aneurysm progression and atherosclerosis, but computing them through computational fluid dynamics is far too slow for routine clinical use. I encode aortic anatomies with a vessel coordinate system that gives point-to-point correspondence across cases, build a statistical shape model to generate physiologically plausible synthetic geometries, simulate them in OpenFOAM, and project the geometric and hemodynamic fields onto spatially aligned maps that a convolutional network can consume. Patient-specific prediction then no longer requires running CFD at inference time.

Rather than replacing the solver, a learned operator supplies an initial guess close to the converged solution, cutting the number of iterations a conventional Navier–Stokes solver needs. The result still comes out of the physics.

Extreme simplification of biological networks as a route to functional discovery, together with the tooling that makes such networks explorable in 3D: layouts, centrality metrics, community detection, geospatial embedding, sparse matrices and immersive VR inspection.

• Deep-learning surrogate models for aortic hemodynamics, predicting wall shear stress, pressure and velocity fields directly from vascular geometry.
• Neural operators used to warm-start Navier–Stokes solvers, aimed at accelerating clinical in-silico trials.
• Design and development of the group's scientific visualization stack, and automation of OpenFOAM pipelines on HPC clusters.
• Mentor of an artist collective producing outreach work around the group's research line.

200 h extracurricular and 180 h curricular internships (graded 10.0/10.0), followed by a formative traineeship from January to October 2025. Supervisor: Prof. Ignacio García-Fernández.

• Built SciBlend, bridging ParaView and Blender for publication-quality visualization of scientific data.
• Developed a Flask + React + Three.js web application to launch parallel OpenFOAM simulations on aortic geometries with a job-queue and semaphore system.

Four years of 3D design, rendering and art direction for more than 100 international clients.

Spanish Ministry of Science, Innovation and Universities, PID2023-148702OB-I00. PI: Prof. Ignacio García-Fernández.

Industry contract (Art. 60 LOSU) with AGROTAN NATURA S.L. PI: Prof. Ignacio García-Fernández.

Industry contract (Art. 60 LOSU) with the Institute of Biomechanics of Valencia. PI: Prof. Ignacio García-Fernández.

Fluid simulation for hemodynamic analysis; mentor of an artist collective creating work for public engagement with the research line. PI: Prof. Ignacio García-Fernández.

GPA 9.3/10. Two Matrícula de Honor distinctions, in Data Visualization and in Geospatial Data Analysis.

Final-year thesis: Deep Learning-Based Prediction of Wall Shear Stress from Thoracic Aorta Geometry, graded 9.8/10. Supervisor: Prof. Ignacio García-Fernández.

J. Marín, I. Marín, I. García-Fernández. The Eurographics Association. 10.2312/ceig.20261030

first authorEurographics

J. Marín, T. M. G. Baptiste, C. Rodero, S. E. Williams, S. A. Niederer, I. García-Fernández. Computers & Graphics 130, 104264. 10.1016/j.cag.2025.104264

first authorJCR indexed

J. Marín, I. Marín. PLOS Computational Biology, PCOMPBIOL-D-26-01766

first authorunder review

Graph Visualization and Analysis with SciGraphs. 50-minute talk, Classroom track.

acceptedspeaker

Shape-Driven Surrogate Modeling of Thoracic Aortic Hemodynamics.

with A. Bayón, F. Martínez-Gil, P. Romero, A. Liberos, M. Lozano, I. García-Fernández

acceptedspeaker

Predicting Hemodynamics from Aortic Geometry.

with A. Bayón, F. Martínez-Gil, P. Romero, M. Lozano, I. García-Fernández

acceptedspeaker

Neural Operators Predicted Warm-Start for Navier–Stokes Solvers in Clinical In-Silico Trials.

speaker

Visualization and Analysis of Graphs with SciGraphs.

with I. Marín and I. García-Fernández

speaker

SciBlend: Advanced Data Visualization Workflows. Recording

speaker

SciBlend: Advanced Data Visualization Workflows within Blender.

speaker

Two posters: SciBlend: Advanced Data Visualization Workflows within Blender, and Deep Learning-Based Prediction of Wall Shear Stress from Thoracic Aorta Geometry.

2 posters

Poster: Deep Learning-Based Prediction of Wall Shear Stress from Thoracic Aorta Geometry.

poster

Python toolkit extending Blender for scientific visualization: import of formats Blender does not read natively (VTK, netCDF, Shapefile), shader and legend generation, coordinate grids, annotation and compositing, with Cycles and real-time EEVEE rendering of large time-varying data.

official Blender ExtensionGPL-3.0

Blender Extensions · GitHub · Zenodo

Import, analysis, styling and rendering of networks inside Blender: graph layouts, centrality metrics, community detection, geospatial data, sparse-matrix input and Geometry Nodes, plus VR immersive visualization and LLM-assisted scene generation.

official Blender Extensionopen source

Blender Extensions · Paper

Computes basic networks from an input graph and a set of seed nodes: a strict graph spanner holding the seeds plus the minimal set of connectors needed to preserve the distances between them. Multi-threaded permutation testing, real-time reporting and an interactive visualization panel, with Windows and Linux builds.

MIT

GitHub