Jeremías Garay L
Assistant Professor @
Universidad Técnica Federico Santa María
"I am an applied mathematician working at the intersection of computational fluid dynamics and inverse problems in hemodynamics. I'm also interested in making computations faster, so their application to clinical problems becomes feasible."
Computational Hemodynamics
Computational hemodynamics studies how blood flows through the body using mathematical models and computer simulations, combining physics, data, and medical imaging to better understand the cardiovascular system. By turning images into predictive tools, it becomes possible to explore quantities that are otherwise difficult to measure and to gain insight into how flow patterns relate to health and disease. The field is rapidly evolving toward faster and more personalized approaches, where models can adapt to patient-specific data and eventually support clinical decisions in a practical and accessible way.
i. Reduced Models for Blood Flow Simulations
Blood flow simulations can become extremely complex, but not every detail is needed to understand what truly matters. I explore simplified models that retain the essential behavior while introducing a small number of meaningful parameters. These parameters can be adjusted to reflect individual patients and may even point toward new ways of characterizing vascular health.
ii. Learning from Data: "Assimilation"
How can we combine simulations with real-world data to improve predictions? I work on methods that continuously adjust models using measurements (such as medical imaging), allowing them to become more accurate and personalized. The goal is to turn simulations into tools that learn and adapt, rather than static predictions.
iii. Neural Operator in Fluid-Structure Interaction
Blood doesn't flow in rigid pipes — arteries move and deform, and this interaction matters. Unfortunately, capturing it accurately is computationally expensive. I investigate how deep learning can accelerate these simulations, making it possible to study complex, patient-specific scenarios that would otherwise be out of reach.
iv. Physics-Informed Neural Networks
Neural networks can solve complex physical problems, but not without challenges. I work on making these methods more reliable and efficient, especially in difficult scenarios involving fluids. This includes understanding where they fail, where they excel, and how to turn them into practical tools for scientific computing.
Curriculum Vitae
Education
Ph.D. in Computational Mathematics
Thesis: On the analysis of inaccurate Phase-Contrast MRI and its assimilation into blood flow models. Advisor: Prof. Cristóbal Bertoglio.
Master's degree in Science, mention Geophysics
Thesis: Flexural model of the subducting plate. Advisor: Prof. Eduardo Contreras.
Bachelor's degree in Science, mention Physics
Employment
Assistant Professor
Faculty member at the Department of Informatics. Conducting research on Physics-Informed Neural Networks (PINNs) and inverse problems in computational hemodynamics, with applications to medical imaging and cardiovascular flows.
Postdoctoral Researcher
Research on the use of deep learning methods for the detection of cardiovascular disease from MRI images, carried out in the Department of Mechanical and Metallurgical Engineering and the Department of Medicine.
Analyst of Seismological Records
Analysis and monitoring of seismological stations in the Chilean territory.
Awards
Best Presentation — The Ridgway Scott Foundation
Best presentation by a postdoctoral researcher during the FEniCS 2025 conference. The prize consisted of 500 USD.
Travel Award — Flax & Teal
Plane tickets for attending the annual FEniCS 2025 conference held in The Netherlands.
ANID — Fondecyt Postdoctorado
Chilean Government. Consisting of salary, equipment and travel budget.
University of Groningen — PhD Scholarship Programme
The Netherlands. Consisting of salary, full tuition, equipment, and travel budget.
Fondecyt Magíster Nacional
For master's studies. Chilean Government. Consisting of salary and full tuition.
International Conference Participation
Languages
Spanish (native), Portuguese (advanced) and English (full proficiency).
Computational Skills
OS: Linux, Windows, Mac · Programming: Python, C
Scientific: MATLAB, FEniCS, PyTorch, Gmsh · Visualization: Paraview, Blender, Unreal Engine 5
Other: Docker, Git
Publications
Garay, J., Nolte, D. & Bertoglio, C. (2026). Duct boundary conditions for incompressible fluid flows: finite element discretizations and parameter estimation in coronary blood flow. Computer Methods in Applied Mechanics and Engineering, 454, 118867.
Garay, J., Dunstan, J., Uribe, S., & Sahli, F. (2024). Physics-informed neural networks for parameter estimation in blood flow models. Computers in Biology and Medicine, 178, 108706.
Garay, J., Nolte, D., Löcke, M., & Bertoglio, C. (2022). Parameter estimation in fluid flow models from aliased velocity measurements. Inverse Problems, 38(9), 095002.
Löcke, M., Garay, J., Franco P., Uribe S., Bertoglio, C. (2022). A comparison of phase unwrapping methods in velocity-encoded MRI for aortic flows. Inverse Problems, 38(9), 095002.
Garay, J., Mella, H., Sotelo, J., Cárcamo, C., Uribe, S., Bertoglio, C., & Mura, J. (2022). Assessment of 4D flow MRI's quality by verifying its Navier–Stokes compatibility. International Journal for Numerical Methods in Biomedical Engineering, e3603.
Garay, J., Ortega, I., Clerc, M. G., & Falcón, C. (2012). Symmetry-induced pinning-depinning transition of a subharmonic wave pattern. Physical Review E, 85(3), 035201.
Teaching
Scientific Computing
Scientific Computing
Material Mechanics
Material Mechanics
Others
Content coming soon.