Cluster Operations Engineer

A cluster that passes acceptance is the starting line, not the finish line.

Our customers get dedicated NVIDIA clusters and a named engineer who knows their workload, not a ticket queue. This is that engineer, for everything above the metal: Kubernetes and Slurm orchestration, the NVIDIA software stack, control plane health, and the customer relationship that turns "here's what we're training" into a cluster configured to do it well.

• →You've operated Kubernetes or Slurm in production and have opinions about scheduler policy.
• →You want to run the NVIDIA toolchain (DCGM, Base Command Manager, NCCL) on the newest hardware NVIDIA makes.
• →You like the customer side: onboarding an ML team, learning what they're building, and shaping the platform around it.
• →You believe a control plane should be boring, monitored, and recoverable.

• →Direct ownership, with no layer between you and the clusters or the customers on them.
• →A fleet spanning GB300 NVL72 to H100, on Quantum-3 XDR InfiniBand, scaling toward 9,552 GPUs.
• →Customers whose training runs you'll know by name.