Loftorbital

ML Runtime Engineer

Toulouse, Occitanie. Posted 4 months ago

The role

You will own model compilation and the performance tooling behind it, turning partner ONNX models into optimized engines that run on flight hardware. You will work across model compilation, runtime internals, quantization and hardware-in-the-loop CI.

What you will do

  • Own model compilation: turn partner ONNX models into optimized engines per accelerator (TensorRT, Hailo, AMD/ROCm) within power and thermal budgets
  • Build the ground-side compile and delivery path, including how a mismatched engine is detected and rejected on-node
  • Build performance tooling: benchmarking, profiling, operator-coverage matrices and budget validation
  • Contribute to the Space Inference Engine (execution providers) and the SDK (model build and packaging)
  • Coordinate with the OBSW / Runtime team

Required qualifications

  • Strong C++ and Python
  • Model compilation with TensorRT (and/or equivalent graph compilers)
  • ONNX Runtime, quantization and inference performance optimization
  • Embedded or edge GPU deployment (NVIDIA Jetson)
  • Benchmarking and profiling / performance tooling
  • CI/CD including hardware-in-the-loop

Nice to have

  • Hailo SDK and/or AMD ROCm compilation (iX10)
  • Writing custom kernels or operator plugins
  • Remote sensing / large-image handling
  • RF / IQ signal data exposure
  • Meson build, Yocto / minimal Debian images
  • Model-weight protection / secure execution

Benefits

  • Equity
  • Up to 35 days of paid time off (vacations and RTT) and flexible working hours
  • Health and life insurance
  • Lunch vouchers
  • Cross-office travel opportunities between San Francisco, Colorado and Toulouse
  • Company and team off-sites and events
  • Relocation assistance to Toulouse when applicable

Before you apply

  • Resume must be attached in English

About the company

Loft provides governments, companies and research institutions with a fast way to deploy missions in orbit, offering end-to-end missions as a service. Founded in 2017, it has flown more than 30 missions and is growing a fleet on track to reach 30 satellites by 2027.

ML Runtime Engineer at Loftorbital | JobZoey