Scientific computing
We build on the numerical methods used to solve thermal, mechanical, and acoustic problems. Our goal is to accelerate computations without changing the physical definition of the problem.
Numerical simulation methods
Our approach integrates with simulation workflows based on geometry, meshing, and numerical discretization. It is designed to work with the tools already used by engineering teams.
Machine learning
Simulations within the same campaign are usually closely related. We use learning methods to identify and reuse useful information from one computation to the next.
Scientific machine learning (SciML)
The methods we develop combine learning models with knowledge from scientific computing. They are designed to remain compatible with the physical and numerical constraints of the problem being solved.
Designed to integrate with existing tools
Our approach complements the simulation tools already in use rather than replacing existing computational workflows.
Learning methods are used to accelerate the campaign without removing the stages required to check and validate the results.