Staff Slurm Cluster & HPC Engineer
Core
Own Slurm cluster architecture, multi-tenant scheduling policy, and cluster reliability on bare-metal and VM-based GPU nodes; lead adoption of the Slinky operator stack to enable shared GPU pools between Slurm and Kubernetes.
Role type
Staff Slurm Cluster & HPC Scheduling Engineer
Builds
Production Slurm clusters on bare metal and VMs, Slinky operator stack, and elastic capacity management between batch training and Kubernetes inference.
Domain
AI infrastructure, High-Performance Computing (HPC), GPU cloud
Deliverable
production ML models
Required skills
Slurm cluster architecture and lifecycle management, topology-aware scheduling for GPU fabrics, multi-tenant scheduling policy, Slinky on Kubernetes implementation, elastic capacity management, container and job runtime operations, cluster health and reliability engineering, automation and infrastructure as code, observability and accounting integration, technical leadership
Preferred skills
Go programming, experience with CoreWeave SUNK or Nebius Soperator
Technologies
Slurm, Slinky, Kubernetes, Terraform, Ansible, InfiniBand, RoCE, NVIDIA DCGM, Prometheus, Grafana, Pyxis, Enroot, containerd, Lustre, GPFS, WEKA, VAST
Responsibilities
Design, deploy, and operate production Slurm clusters on bare metal and VMs; model physical fabric for rail-optimized InfiniBand/RoCE and NVLink domains; own account/association tree, partitions, QOS, fairshare, preemption, and reservations; lead implementation of Slinky slurm-operator and evaluate slurm-bridge; shift GPU nodes between batch training queues and Kubernetes inference capacity; operate Pyxis/Enroot and OCI/containerd job paths; build passive and active health-check systems for GPU clusters; deliver clusters through Terraform/Ansible and golden images; instrument queue wait time, allocation efficiency, and job failure taxonomy; write runbooks, onboard enterprise customers, and mentor platform engineers
Seniority
Staff, hands-on IC with technical leadership
