Infrastructure Engineer (ML Compute) - Paris
This role is about keeping a fleet of GPUs healthy, busy and well placed so research and product teams can train and serve models. A strong candidate has run GPU workloads on Kubernetes, can trace faults down through Linux, drivers, networking and storage, and writes solid Python. Experience as an SRE or infrastructure engineer helps, and ML research background is not required.
How it fits the AI era
AI models are the main customer here, so the work centers on feeding them reliable compute, and tools that detect and drain bad nodes can automate much of the routine upkeep. What stays human is the judgment: deciding how to trade capacity between training and serving, handling messy failures that cross hardware and software, and negotiating with cloud providers. Good infrastructure choices directly decide how fast a team can try new ideas.
How it builds AI experience
Worth showing
- GPU workloads on Kubernetes
- Linux and driver debugging
- Strong Python
- Incident handling and node repair
This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.