Job Description & Details
This is a heavy-hitting architecture role focused entirely on scaling out GPU-as-a-Service infrastructure, meaning you will be deep in the weeds of multi-tenant orchestration and high-performance computing clusters. If you like solving the messy problems of GPU sharing, node scheduling, and keeping enterprise AI workloads from stepping on each other's toes, this gig is right in your wheelhouse.
What You'll Actually Be Doing
You will spend your days designing and governing the end-to-end GPU platform that enterprise AI teams rely on. Your core responsibilities involve figuring out how to slice up expensive NVIDIA hardware efficiently using MIG, making sure Slurm or Kueue schedules jobs without bottlenecking, and building out self-service automation so customers can provision clusters without bugging engineering. You are also going to be the technical bridge between enterprise clients and distributed offshore engineering teams, translating fuzzy AI requirements into solid Kubernetes blueprints.
The Core Tech Stack
You need deep, production-level scars with Kubernetes and OpenShift, paired with hands-on mastery of the NVIDIA GPU Operator, Network Operator, and MIG Manager. Beyond container plumbing, you must know your way around advanced workload schedulers like Slurm, Volcano, or Kueue because default Kubernetes scheduling just won't cut it for massive AI training runs. Toss in heavy Terraform, Helm, and GitOps experience for automation, and you have the complete toolkit required to survive here.
Interview Expectations
The hiring manager is going to grill you on failure modes in multi-tenant GPU environments, likely asking how you handle noisy neighbors monopolizing VRAM or network fabric during a distributed training run. They want to see that you understand the exact limits of NVIDIA Multi-Instance GPU (MIG) versus time-slicing. Be ready to whiteboard a complete provisioning pipeline from a customer clicking 'deploy' in a service catalog to pods scheduling onto specialized GPU nodes.
Application Advice
Your resume needs to scream GPU platform engineering, not just general DevOps or web infrastructure. Make sure keywords like GPU-as-a-Service, NVIDIA GPU Operator, MIG, Slurm, Kueue, and Multi-Tenancy are front and center in your recent experience. Don't just list tools; highlight the scale of the clusters you've managed and specific multi-tenant isolation challenges you've solved to get past the automated screen.