Software Engineer (CUDA)

Crossing Hurdles; La Réunion; Sur place; EUR 60 000 - 90 000; Plein temps

Résumé du poste

Crossing Hurdles is seeking a results-driven GPU optimization engineer to analyze, profile, and optimize CUDA kernels across modern GPU architectures. You will refactor C++ and CUDA code, implement shader logic with GLSL/WebGPU, and document performance gains for stakeholders.

You will collaborate with teams, identify bottlenecks, and propose strategies to improve efficiency, with a strong emphasis on cross-hardware performance and clear reporting.

Qualifications

• CUDA programming expertise with proven performance-tuning history.
• Advanced C++ in HPC environments.
• Experience with GLSL and WebGPU shader development.
• Experience with GPU profilers (Nsight/Visual Profiler).
• Strong analytics to evaluate kernel performance across GPUs.

Responsabilités

• Analyze, profile, and optimize GPU kernels using CUDA and relevant profiling tools.
• Collaborate with stakeholders to identify kernel bottlenecks and propose optimization strategies.
• Refactor C++ and CUDA codebases for improved efficiency across GPU architectures.
• Implement shader logic and graphics workflows using GLSL and WebGPU.
• Document findings, optimization steps, and performance improvements with clear reports.
• Contribute expertise to design discussions and evaluate new GPU-based approaches.

Connaissances

CUDA programming

C++ development

GPU profiling

Shader programming GLSL

WebGPU experience

Analytical skills

Outils

Nsight

Visual Profiler

Description du poste

• Analyze, profile, and optimize GPU kernels using CUDA and relevant profiling tools.
• Collaborate with stakeholders to identify kernel bottlenecks and propose optimization strategies.
• Refactor C++ and CUDA codebases for improved efficiency across GPU architectures.
• Implement shader logic and graphics workflows using GLSL and WebGPU.
• Document findings, optimization steps, and performance improvements with clear reports.
• Contribute expertise to design discussions and evaluate new GPU-based approaches.

Requirements

• Demonstrated expertise in CUDA programming with a strong track record in performance-tuning.
• Advanced C++ development skills in high-performance computing environments.
• Hands‑on experience with GLSL and WebGPU for shader development.
• Proficiency using GPU profilers like Nsight or Visual Profiler for optimization.
• Strong analytical abilities to evaluate kernel performance across hardware generations.