Operations Engineer, Fleet Reliability

About the role

As we bring up owned clusters alongside our cloud capacity, we're hiring Operations Engineers to keep the fleet alive. This is a hands-on role. You're first in line when nodes go bad, GPUs throw ECC errors, IB links flap, or a rack stops responding. You'll provision new nodes, validate them, ship them to production, and troubleshoot whatever entropy throws at them. You'll be on-call. You'll be in the weeds.

You're a fit if you've:

• Administered Linux Systems in the critical path before
• Troubleshooted GPU node issues: NVLink, NCCL, IB, driver and firmware bugs
• Has experience in observability systems like Grafana and Prometheus
• Scripted your way out of repetitive work (bash, python, go, whatever)

Who you are:

• Curious. You don't accept "it's flaky" as a root cause
• Comfortable with ambiguity. The runbook doesn't exist yet for half of what you'll do
• On-call doesn't scare you
• You'd rather automate a problem than fix it twice

Responsibilities:

• Provision, validate, and triage GPU nodes across B300, H200, and H100 clusters
• Troubleshoot hardware and software issues across compute, network, and storage
• Monitor fleet health, take remediation action, push fixes upstream when needed
• Write the runbooks. Improve the ones that exist. Delete the ones that don't work