Computational Physicist · Research Software Engineer (HPC & Numerical Methods) --- Waiting for my PhD defense :)
Max Planck Institute for Gravitational Physics (Albert Einstein Institute) and
IMAPP @ Radboud University Nijmegen (supervisors: Erik Schnetter, Badri Krishnan).
I turn high-level physics into scalable, reliable simulation software, from numerical method design to HPC automation and performance.
📍 Nijmegen, NL & Leipzig, DE 🌐 https://svretina.github.io/ ✉️ stamatis.vretinaris@aei.mpg.de · svretina@gmail.com
High-order numerical operators for spherical and curvilinear coordinates, enabling symmetry-reduced simulations with theoretical O(N^2) speedups and memory reductions over full 3D Cartesian discretizations.
Stack: Julia · Numerical Analysis · Performance Optimization
🔗 https://github.com/svretina/SphericalSBPOperators.jl
📄 arXiv: https://arxiv.org/abs/2606.05155
Numerical framework for scalar self force in general relativity using Finite Volume Method.
Stack: Julia · FVM
🔗 https://github.com/svretina/ScalarWaveFVM
📄 arXiv: https://arxiv.org/abs/2606.06487
Python-based automation pipeline that processed 100,000+ simulations on HPC systems (HTCondor), with robust job orchestration and post-processing.
Stack: Python · HTCondor · Bash/Linux · Git (CI/CD)
🔗 https://github.com/svretina/Ektome
An AVX-optimized Julia library for fast Tanh-Sinh quadrature. Achieved up to 62× speedup in 1D, 35,700× in 2D, and successful microsecond-scale convergence in 3D boundary-singular integrals where competing libraries failed to meet tolerance.
Stack: Julia · Numerical Analysis · Performance Optimization
🔗 https://github.com/svretina/FastTanhSinhQuadrature.jl
📄 JOSS: https://joss.theoj.org/papers/10.21105/joss.10076
Improved parameter estimation for gravitational-wave models by ~10× using physics-informed priors and machine learning.
Stack: Python · Bayesian inference · ML
🔗 https://github.com/svretina/PythiaBNS
📄 arXiv: https://arxiv.org/abs/2501.11518 · PRD: https://journals.aps.org/prd/abstract/10.1103/g1qs-j74x
- High-Performance Computing (HPC): Orchestrating large-scale Slurm/HTCondor workflows, automating simulation pipelines, and conducting cluster scaling studies.
- Performance Engineering: Profiling and benchmarking (e.g., LIKWID) for hardware-aware optimization and high-efficiency software development.
- Research Software Engineering: Architecting maintainable, production-ready codebases with Git, robust CI/CD pipelines, and comprehensive documentation.
- Scientific Computing: Implementing numerical methods (FVM/FDM), high-order methods, and rigorous stability/convergence analysis.
- Languages: Python · Julia · C/C++ · Fortran · Bash · Mathematica
- Infrastructure & Tools: Linux toolchains · Slurm · HTCondor · OpenMP · Git · CI/CD
- Methods & Algorithms: Bayesian Inference · Monte Carlo · Scientific ML · Finite Volume/Difference