Mathias Louboutin
Scientific computing · Open source · Inverse problems
Turning complex science into usable software
I’m Mathias Louboutin, a scientific researcher and software developer building high-performance tools for wave-based imaging, optimization, and machine learning.
Senior Solution Architect at Devito Codes · Python, Julia, HPC, and cloud computing
Open to research collaborations
What I work on
Research at the intersection of mathematics and computing
I develop computational methods and open-source software that make demanding scientific problems easier to express, scale, and solve—from a laptop to the cloud.
01
Inverse problems
PDE-constrained optimization for seismic, photoacoustic, and ultrasound imaging, with an emphasis on practical large-scale solvers.
02
High-performance computing
Task-parallel methods for exascale simulations, large scientific datasets, and performance-portable numerical workloads.
03
Scientific software
Domain-specific languages and symbolic APIs that let researchers describe complex models without sacrificing computational performance.
04
Machine learning
Randomized linear algebra and memory-efficient methods that connect modern learning systems with physics-based computation.
Featured work
Open tools for ambitious science
Python · Compiler · HPC
Devito
A symbolic finite-difference domain-specific language and just-in-time compiler that generates optimized C for CPUs and GPUs.
Julia · Linear algebra · Imaging
JUDI
A matrix-free framework for PDE-constrained optimization that combines a high-level Julia interface with Devito’s computational performance.
Learning · Efficiency · Open source
Memory-efficient ML
Methods and software for scalable learning, including randomized gradients, reversible networks, and low-memory scientific workflows.
Let’s collaborate
Have a difficult computational problem?
I enjoy working with researchers and engineering teams on scientific software, large-scale simulation, imaging, and high-performance computing.