Come work on a problem that has not yet been solved.

We are a small technical team with roots in public research. Everyone sees their work make its way into the product, and no one is far removed from the actual computations.

Write to us contact\@cartanlabs.fr

How we work

Make decisions from results

When several approaches are possible, we compare them through reproducible experiments and against criteria defined in advance.

Make negative results visible

Experiments that fail or produce insufficient results are documented just like positive ones. They help guide subsequent development.

Understand the methods

We try to understand the source of a numerical behavior before changing a method or its parameters.

Ensure reproducibility

Simulation campaigns, experiments, and code are documented so that another member of the team can review, verify, and reproduce them.

A small team, on subjects that cannot be delegated.

We work on real campaigns, with real meshes and real computation times. There is no team that thinks and team that executes: whoever designs a method watches it run, fail, then pass.

The subject sits at the boundary of several disciplines, so we explain a great deal to each other. That is the best part of the job: nobody knows everything, and everyone learns from someone every week.

The profiles we are looking for

Numerical analysis

Finite elements, iterative methods, preconditioning, model reduction. You know why a solver stalls.

Learning and physics

SciML, constrained learning, transfer across tasks. You are cautious about models whose outputs cannot be verified.

Scientific software engineering

High-performance computing, traceability, reproducibility. You care about being able to reproduce a result exactly.

Interested in talking?

Tell us what you can do and what kind of problem you would like to work on. We do not require you to fit a predefined job description: we read every application and reply.

Write to us