Numerical methods · Verification · Custom solvers
Simulation code that is right, fast and verified.
I verify, fix, parallelise and build simulation codes for R&D teams. From in-house codes to modules for commercial software, you get results you can trust, faster computations and the evidence your customers and regulators ask for.
- PhD, École des Ponts – Polytechnic Institute of Paris
- Industrial PhD with CEA
- Researcher at the University of Basel
Services
From code review to custom solvers
I work across the whole life cycle of your simulation codes. Every mission has a written scope and a tailored quote.
Code diagnostic
Correctness, performance and maintainability of your code, with prioritised recommendations.
Custom solver or module
The solver or module your code lacks, whether commercial software (Abaqus, Ansys, COMSOL…) or your in-house code.
Verification and validation
Quantified, reproducible proof that your code is right, up to regulatory files.
Code modernisation and industrialisation
From a research prototype or legacy code to a modern, verified tool, parallelised on CPU and GPU.
Performance and parallelisation
CPU and GPU parallelisation (MPI, OpenMP, OpenACC), choice of solvers, profiling, scaling.
AI and surrogate models
Reference computations and an audit of where your AI models remain valid.
Second opinion on a simulation
An independent view on how far a computation can be trusted: convergence, assumptions, errors.
Fractional numerical expert
A few days a month of senior numerical judgement, without hiring.
A specific request?
Every project is different. If your need does not fit any of these services, describe it to me and I will tell you frankly whether I can help.
Expertise
Broad knowledge of modelling
A simulation code is only right if the model is well understood. I know the models behind your computations, alone or coupled.
Navier–Stokes equations
Incompressible flow, steady or unsteady.
Solid mechanics
Elasticity, plasticity, large deformations, contact, constitutive laws.
Heat equation
Conduction, internal sources, heat exchange at walls.
Wave equation
Acoustics, ultrasound, non-destructive testing.
Maxwell’s equations
Electromagnetics in the time or frequency domain.
Advection–diffusion–reaction
Species transport, pollutant dispersion, chemical kinetics.
Black–Scholes equation
Option pricing and risk models, by finite differences, finite elements or Monte Carlo.
Nonlinear problems
Newton methods, continuation, convergence control.
Multiphysics coupling
Heat and flow, fluid and structure, electromagnetics and heat.
Why a methods specialist
What you get from a numerical analyst
Independent of software vendors
I sell no licences. I recommend what works for your problem, open-source software included.
Verified results
Every deliverable includes quantified verification you can rerun: tests, convergence orders, scripts.
A firm quote before work starts
Written scope, fixed-price quote by milestone, 30% deposit. You know the price before work starts.
Your code is yours
You own the deliverables. My generic tools come with a free, perpetual licence of use.
How a mission runs
Three simple steps
A free diagnostic
30 minutes to understand your problem.
A written quote
Scope, timeline and price agreed upfront.
Delivery
Code, tests and report, walked through with your team.
Who I work with
Concrete examples, sector by sector
Abaqus, Ansys or COMSOL users
Write and verify the material law, boundary condition or coupling your software is missing.
Instrumentation and sensors
Predict how fouling affects an ultrasonic flow meter, or how an inductive sensor responds to a metal part.
Simulation software vendors
Add a physical model to your solver, and prove with test cases that it converges at the expected order.
Industrial teams with legacy code
Modernise a twenty-year-old Fortran code, parallelise it on CPU and GPU, and prove it still gives the same results.
Pharma and chemicals
Model mixing in a bioreactor or separation in a chromatography column, and verify the model before it replaces experiments.
Medical devices
Show convergence and quantify uncertainty for an implant or blood-flow simulation in a regulatory file.
Energy, nuclear and geosciences
Verify a thermal-hydraulics code before qualification, or simulate seismic wave propagation through complex ground.
Quantitative modelling
Validate a PDE or Monte Carlo model, and cut its run time substantially.
Deep-tech start-ups
Turn a lab prototype into a reliable, documented solver before a funding round.
Let’s talk about your code.
A free 30-minute diagnostic is enough to see whether I can help, and at what price.