Curriculum vitae

Sebastien Dubois

ENS Paris-Saclay Normalien & Scientific Software Engineer

Impact & leadership

Experience

Palaiseau, France · Sep 2024 – Present

École Polytechnique — Center of Applied Mathematics (X/CMAP)

Lead the Numpex modernisation of the Samurai solver with coordinated releases for research users.
Operate regression and performance testing with ReFrame and reproducible multi-node benchmarks.
Built internal CLIs, maintained build/release pipelines, and established incident runbooks.
Partner with ONERA and CMAP scientists to translate research code into maintainable components.

Palaiseau, France · Sep 2022 – May 2024

Office National d'Études et de Recherches Aérospatiales (ONERA)

Contributed C++ features to the elsA industrial CFD solver improving MPI/OpenMP workflows.
Integrated Fortran linear-algebra kernels into research prototypes with validation alongside scientists.
Ran 800+ rank Slurm benchmarks and produced tuning guidance for engineering teams.

Paris-Saclay, France · Apr 2021 – Aug 2021

Commissariat à l'Énergie Atomique et aux Énergies Alternatives (CEA)

Optimised GPU pipelines for Material Point Method simulations to achieve real-time performance.
Authored comparative studies on fidelity and stability adopted for industrial rollout.

Selected work

2024

Built an interactive canvas-based simulation that visualises chaotic motion with persistent trails.
Runtime controls expose the underlying physics (masses, lengths, gravity, launch angles) and recalibrate the system instantly.
Documented the equations of motion to help visitors understand the transition from deterministic rules to unpredictable patterns.

Aug 2025 – Present

Delivering a hosted platform with FastAPI endpoints, CLI tools, and poetry-based packaging.
Redis orchestrator managing AWS/GCP compute, hierarchical profiling, and S3-compatible storage.
Postgres and vector DB persistence layered with Sentry instrumentation and CI/CD automation.

Sparse neural network library with MPI backend using Kokkos/Kokkos-Kernels for CUDA portability.
Implemented Adam optimiser, pruning strategies, and distributed training workflows.

U-Net predicting pressure and velocity fields from CFD snapshots on CUDA backends.
C++ simulation pipeline with Kokkos generating datasets via hierarchical parallelism.

Academic path

Normalien Élève — Computer Science & HPC

MSc High Performance Computing & Simulation (CHPS)

MSc Modelling & Simulation in Structural Mechanics (MS2SC)

Capabilities

Languages
French (Native)English (Professional)PythonCC++Fortran
Platforms & Data
SlurmDockerGCP (Google Cloud)PostgresVector databases
HPC
ProfilersParallelizationVectorizationDistributed computingAlgorithms
Quality & Tooling
pytestGitHub ActionsSentryReFrameCMake
Collaboration
Partnering with scientistsScientific communicationDesign notesCode reviews
Soft skills
TeamworkSystems thinkingContinuous learning