Responsibilities
- Convert partner-provided ONNX models into optimized inference engines tailored to specific hardware accelerators, including TensorRT, Hailo, and AMD/ROCm, while adhering to power and thermal constraints
- Develop and maintain the ground-based pipeline for model compilation and deployment, ensuring mechanisms are in place to detect and reject incompatible engine builds on the target node
- Create performance analysis tools for benchmarking, profiling, operator coverage assessment, and verification of system budget compliance
- Support development of the Space Inference Engine’s execution providers and SDK components related to model build and packaging, while collaborating closely with the OBSW and Runtime teams
Work Arrangement
Hybrid
Other
- Resume must be submitted in English
- Relocation support provided for positions based in Toulouse when required
- Flexible working hours offered
- Opportunities for travel across offices in San Francisco, Colorado, and Toulouse
