Production Engineer - Large-Scale GPU Compute Fleet

Fluidstack; California; On-site; USD 180,000 - 250,000; Full time

Benefits offered by this job

Stock options

Health, dental, and vision insurance

Generous PTO

Job summary

Fluidstack is hiring for a Production Engineering role focused on building and operating one of the world’s largest GPU fleets. You will own the health, observability, and automation of production compute across Kubernetes and bare metal, driving incident discipline and scalable repair pipelines.

You will qualify GPUs, manage fleet telemetry, and expand the qualification platform for new generations, with a culture of extreme ownership and rapid iteration.

Qualifications

• Toil is treated as a bug; eliminate manual repair steps.
• Strong hardware intuition, including firmware/silicon failure modes.
• Able to navigate ambiguity, map complexity, and explain it clearly.
• Rapid learning with unfamiliar domains; value over specific past expertise.
• Experience shipping production automation and working with AI tooling.

Responsibilities

• Own compute fleet health end-to-end; build metrics, alerts, and a unified view.
• Turn deployment/repair into a scalable automation pipeline; avoid one-off scripts.
• Design and expand GPU qualification platform for new generations.
• Own Redfish and BMC tooling; telemetry, logs, and fleet-access layer.
• Ensure reliability, scalability, and operation of the compute fleet at scale.

Skills

Kubernetes

GPU hardware

Observability

Automation

Python

Go

Redfish

Tools

Prometheus

Grafana

Temporal

Cadence

Job description

Fluidstack is hiring for a Production Engineering role focused on building and operating one of the world’s largest GPU fleets. You will own the health, observability, and automation of production compute across Kubernetes and bare metal, driving incident discipline and scalable repair pipelines.

You will qualify GPUs, manage fleet telemetry, and expand the qualification platform for new generations, with a culture of extreme ownership and rapid iteration.