Platform Engineer – GPU Infrastructure (Kubernetes / Slurm / Bare Metal)
📍 Sweden, fully remote
Build a GPU platform from bare metal up.
A well-funded, fast-growing infrastructure company is building out a large-scale GPU platform that AI teams will train and run models on. They want engineers who’ve run GPU fleets in production and want to build the platform, not just operate it.
You’d own the platform layer from the metal up:
- Provisioning and lifecycle for bare-metal GPU nodes (drivers, firmware, CUDA, burn-in, decommissioning)
- GPU scheduling and multi-tenancy with Kubernetes (GPU Operator, MIG) and/or Slurm
- High-performance fabrics for distributed training (InfiniBand/RoCE, NCCL tuning)
- Fleet health: detecting bad GPUs and links, plus automated drain-and-recover (DCGM, Prometheus)
- Control-plane services, mainly in Go
You’ve probably:
- Kept a GPU or HPC cluster alive in production, and been the one paged when nodes went down
- Written real software around infrastructure in Go, Python, Rust or C++, not just config
- Run Kubernetes and/or Slurm at scale
You don’t need every item on the list. Deep experience in a couple of these areas plus the instinct to learn the rest is what matters.
Why it’s interesting: greenfield platform work, real hardware at scale, a small senior team, and a lot of ownership early on.
You’ll need the existing right to work in Sweden and must pass a background check.
