Syver D. Agdestein↗︎

A little about me

Curiosity across scales

I’m Syver Døving Agdestein, a postdoctoral researcher in scientific computing at CWI in Amsterdam.

Portrait of Syver Døving Agdestein

I work on probabilistic simulation of turbulence, closure models, and differentiable programming in Julia. The central question: how do we account for the detail that a practical simulation cannot resolve?

I did my PhD at CWI on data-driven closure models for large-eddy simulation, treating the numerical discretization as part of the physics. Eindhoven University of Technology awarded the degree cum laude, and the thesis received the best thesis prize of the Dutch–Flemish Scientific Computing Society.

Away from research, I’m drawn to history, languages, and the piano. I’ve lived in Asker, Gembloux, Toulouse, Saint Petersburg, Paris, and now Amsterdam.

01Digging in

Employment

  1. 2026–Present

    Postdoctoral researcher

    Scientific Computing group, CWI ↗︎ · Amsterdam, the Netherlands
    Generative machine-learning closure models for large-eddy simulation.

  2. 2021–2026

    PhD researcher

    Scientific Computing group, CWI ↗︎ · Amsterdam, the Netherlands
    A mathematical framework for discretization, filtering, and structure-preserving machine learning for large-eddy simulation of incompressible turbulence.

  3. 2020–2021

    Software Engineer

    INRIA / École Polytechnique ↗︎ · Palaiseau, France
    Eigenfunction-based reduced-order modeling for direct simulation of diffusion MRI of brain tissue · 6 months

Earlier research internships

  1. 2020

    Research intern

    IRT Saint Exupéry ↗︎ · Toulouse, France
    Mixture-of-experts machine-learning models for multidisciplinary optimization of aircraft design · 6 months

  2. 2019

    Research intern

    Toulouse Mathematics Institute, CNRS ↗︎ · Toulouse, France
    Data-driven learning of nonlinear diffusive-wave equations for river hydraulics from satellite data · 8 weeks

  3. 2018, 2019

    Research intern

    Simula Research Laboratory ↗︎ · Oslo, Norway
    Finite-element solvers for blood flow in arterial networks and arterial-wall deformation, including Artery.FE, a FEniCS implementation of the 1D blood-flow equations · 2 × 6 weeks

02Learning the trade

Education

  1. 2021–2026

    PhD in applied mathematicscum laude

    Eindhoven University of Technology ↗︎ · the Netherlands
    Awarded May 2026. Thesis: Data-driven discrete closure models for large-eddy simulation of incompressible turbulence (research carried out at CWI, Amsterdam). Supervised by Benjamin Sanderse (CWI) and Roel Verstappen (University of Groningen).

  2. 2015–2020

    Master of applied science and technology / Diplôme d'Ingénieur

    INSA Toulouse ↗︎ · France
    Integrated Bachelor’s and Master’s degree, specializing in mathematical modeling.

    2018–2019 · Exchange year

    Peter the Great St. Petersburg Polytechnic University · Saint Petersburg, Russia
    Theoretical mechanics and applied mathematics

  3. 2012–2015

    Asker Upper Secondary School

    Norway

    2013–2014 · Exchange year

    Collège Saint-Guibert · Gembloux, Belgium

03Crossing the line

Distinctions

  • 2026
    Dutch–Flemish Scientific Computing Society (SCS)Best thesis prizeFor the PhD thesis on data-driven discrete closure models for large-eddy simulation.
  • 2026
    Royal Dutch Mathematical Society (KWG)KWG PhD prize finalistOne of five theses shortlisted nationally, presented at the KWG afternoon session in Amsterdam.
  • 2024
    Dutch–Flemish Scientific Computing Society (SCS)Best poster prizeAwarded at the Woudschoten Conference in Zeist.

04Passing it on

Teaching & supervision

Master’s thesis supervision

  • 2026 · Eindhoven University of TechnologyLucas RonckersProbabilistic turbulence modeling with ideal large eddy simulation: Bayesian inverse filtering and flow matching
  • 2023 · University of AmsterdamViviane DesgrangeAn inverse problem approach for closure modelling

Lectures at schools and masterclasses

  • October 2023 · CWILearning neural closure models for fluid flowsAutumn School on Scientific Machine Learning
  • May 2022 · CWILearning physics from dataMasterclass on Machine Learning for Inverse Problems: A Bayesian Perspective

05Lending a hand

Service & outreach

06Tools for the road

Skills & languages

Research

  • Numerical analysis
  • Partial differential equations
  • Finite-element and finite-volume methods
  • Spectral methods
  • Reduced-order modeling
  • Adjoint methods
  • Turbulence and large-eddy simulation
  • Structure-preserving machine learning
  • Generative modeling
  • Diffusion-MRI simulation

Computing

  • Julia
  • Python
  • MATLAB
  • High-performance computing with Slurm
  • Differentiable programming
  • Multi-GPU kernel programming
  • Sparse linear algebra
  • Tetrahedral meshing
  • Git
  • Continuous integration
  • Agentic workflows

Languages

  • HeiNorwegianNative
  • HelloEnglishFluent
  • BonjourFrenchFluent
  • ПриветRussianBasic
  • HalloDutchBasic
Find something interesting