HPC Benchmarking & Profiling Engineer M/F

Numpex; France; Sur place; EUR 60 000 - 90 000; Plein temps

Résumé du poste

NumPEx is seeking an HPC Benchmarking & Profiling Engineer to join the Exa-DI team and contribute to automated benchmarking toolchains and reproducible profiling workflows.

Based in France, the role covers performance measurement across CPU, GPU, and many-core architectures, with mentoring and cross-disciplinary collaboration. Starting June 2026, it spans a 36-month project within the Exa-DI initiative.

Qualifications

• Master’s degree, engineering degree, or PhD in computer science or related scientific computing field.
• Proficient in Python and C/C++, with in-depth knowledge of parallel programming (GPU, MPI).
• Familiar with Git, GitHub/GitLab, CMake/CTest, Docker, Spack, Guix, etc.
• At least one significant experience in HPC benchmarking & profiling, with tool/infrastructure and metrics.
• Pragmatic, initiative-taking, with strong analytical and cross-disciplinary collaboration skills.

Responsabilités

• Build and maintain benchmarking toolchains and CI-integrated HPC benchmarks.
• Design profiling data collection workflows with traceability and bottleneck analysis.
• Define evaluation metrics for performance, scalability, portability, and efficiency across CPU, GPU, and distributed clusters.
• Develop benchmark scenarios and reference workloads for Exa-DI mini-applications.
• Lead benchmarking campaigns, ensuring reproducibility and architectural comparability.
• Identify bottlenecks and advise on appropriate parallel programming models and runtimes.
• Lead knowledge transfer and training on benchmarking methodology for developers and researchers.
• Contribute benchmark insights to software co-design and guide technical decisions.

Connaissances

Python

C/C++

Parallel programming (GPU, MPI)

CI/CD tools

Git/GitHub/GitLab

CMake/CTest

Docker

Spack

Guix

CUDA/HIP/SYCL

Formation

Master’s degree

Engineering degree

PhD

Outils

Git

GitHub/GitLab

CMake/CTest

Docker

Spack

Guix

CUDA

HIP

SYCL

Description du poste

The successful candidate will join the NumPEx Exa-DI project.

General informations

The successful candidate will join the NumPEx Exa-DI project at:

• Duration: 3 years (36 months)
• Starting date: June, 2026

Context

Launched in 2023 for a duration of six years, the NumPEx PEPR Research Program contributes to the design and development of numerical methods and software components that will equip future European exascale and post-exascale machines. NumPEx also aims to support the scientific and industrial community in fully leveraging the capabilities and potential of these new architectures. The application domains include, among others: meteorology, climatology, aeronautics, automotive, astrophysics, high-energy physics, materials science, energy production and management, biology, and healthcare.

Within NumPEx, the Exa-DI team works hand-in-hand with the application community to:

• Identify and formalize key algorithmic and communication patterns encountered in exascale applications
• Specify mini-applications that capture their core technical challenges
• Develop and deliver these mini-applications, based on the NumPEx software stack, as reference implementations for research, co-design, and performance evaluation
• Evaluate the performance and portability of different mini-app implementation on large scale facilities and hardware architectures

To strengthen its benchmark and profiling expertise, Exa-DI is creating a dedicated HPC Benchmarking & Profiling Engineer position—a role designed to drive performance measurement rigor, automation, and reproducibility for the team’s mini-applications.

Mission

As our HPC Benchmarking & Profiling Engineer, you will be responsible for structuring, executing, and scaling the team’s benchmarking efforts. In this role, you will:

• Build and maintain the benchmarking toolchain, including frameworks and infrastructure for automated, reproducible, and CI-integrated HPC benchmarks
• Design and implement profiling data collection workflows, ensuring traceability and systematic bottleneck characterization
• Define and standardize evaluation metrics covering performance, scalability, portability, and efficiency across architectures (CPU, GPU, many-core, distributed clusters…)
• Develop benchmark scenarios and reference workloads for Exa-DI mini-applications, enabling statistically significant cross-platform comparisons
• Lead benchmarking campaigns, ensuring result relevance, reproducibility, and architectural comparability
• Identify performance and portability bottlenecks, and advise on the most effective use on parallel programming models (e.g., MPI, OpenMP, CUDA, HIP, SYCL…), runtimes (e.g., StarPU), and abstraction layers (e.g., Kokkos)
• Drive knowledge transfer, through mentoring, internal support, and training on benchmarking methodology and profiling tools, for both the development team and the application community
• Contribute benchmark insights to software co-design, helping guide technical decisions using measurable evidence

You will also participate in the team’s Agile practices, including: process improvements, project planning and tracking, progress reviews and demonstrations, and coordination with other development and research teams.

Required Skills

You hold a master’s degree, an engineering degree, or a PhD in computer science or another field related to scientific computing.

You are proficient in multiple programming languages (Python, C/C++), ideally with in-depth knowledge of parallel programming (GPU, multi-threading, MPI, etc.). You are familiar with standard collaborative development tools: Git, Gitlab/GitHub, CMake/CTest, Docker, Spack, Guix, etc.

You have at least one significant experience in HPC benchmarking & profiling, with demonstrable contributions in both the engineering aspect (tools & infrastructure) and the scientific aspect (defining meaningful metrics).

You are pragmatic and take initiative. Your analytical skills and ability to step back allow you to confidently tackle complex problems with multiple constraints (deadlines, major technical challenges).

You enjoy teamwork and have a strong interest in interdisciplinary collaborations involving multiple stakeholders at the intersection of applied mathematics, computer science, and physics applications. You have excellent written and verbal communication skills, both in French and English.