Compute Engineer, Deployment

Fluidstack Job Opportunity

Fluidstack exists to make humanity more free. For most of human history, you farmed or you starved. Technology gave people more time for the things they wanted to do, instead of things they had to do. Powerful AI will be the biggest lever for human choice we've ever built - but only if models are aligned with what humanity actually wants.

There are groups building AI who don't share these goals. Whoever deploys frontier compute infrastructure fastest will decide whether AI expands human freedom or shrinks it.

We're singularly focused on delivering 10 to 100s of GWs of compute faster than anyone else, rethinking every layer of the stack. We acquire power, design and build data centers, and operate them - with teams spanning hardware and software. Speed and scale are our key differentiators. Come be a part of building civilization-scale infrastructure for AI.

We hire people who care deeply about this problem space. If that is you, please apply!

• Extreme ownership. Full autonomy. Own things end to end often taking on scope outside your core role without being asked to get things done.
• Velocity. We drive everything forward as fast as possible.
• First principles. Challenge every assumption. Zero analogy thinking, no egos, the best idea wins.
• Love of the game. The frontier of AI is the most interesting problem of our time. We put in long hours at high intensity to push the frontier forward.

Examples of key problems the team is working on:

• Bring gigawatts of accelerators from first power-on to production. Facility availability to ready-for-service across thousands of racks per site, with a new data hall landing every few weeks.
• Make rack qualification faster than the fleet grows. Firmware baselines, burn-in, and cluster validation proven on every rack before a customer workload touches it, at a pace that never becomes the critical path.
• Scale by tooling, not headcount. Deployed megawatts grow severalfold next year while the team stays near-flat, because anything done twice by hand becomes software.

• Own compute turn-up from facility availability to ready-for-service: the stretch after the network hands off and before customers run workloads.
• Qualify racks at scale: establish firmware baselines, configure BMC and BIOS, run burn-in, and validate at node and cluster level across hundreds of racks per site on GPU and custom accelerator platforms.
• Drive qualification through the base-management Kubernetes platform and provisioning stack (discovery, imaging, firmware updates, shared services), burning down qual queues with tooling rather than manual runs.
• Triage hardware failures found in qualification: isolate to component, drive RMA and vendor escalation, and feed failure patterns back into the qual gates.
• Run turn-up remotely by default, with on-site pulses of roughly a week per data hall as new halls reach facility availability, plus occasional overlapping-site weeks.
• Partner with network deployment, ICT, data center operations, and hardware teams during turn-up windows, and support incident response on freshly-live capacity.

The below is a starting point. We always make space for exceptional people, so if you don't fit this role exactly, tell us where you would.

• You've brought up server or GPU fleets at scale, hundreds of nodes or more, and taken them all the way to production.
• You work deep in Linux and out-of-band management: BMC, IPMI, and Redfish are daily tools for you, not occasional lookups.
• You've automated hardware workflows in Python or Go rather than clicking through them, and the second time you do anything by hand you turn it into software.
• You've worked physically in data halls, racking, cabling, and swapping components, and you're just as effective acting as remote hands or directing them.
• You triage failures methodically across hardware, firmware, and software, isolating the fault to a component before reaching for a fix.
• You travel for turn-up windows when a new data hall comes online.
• Bonus:
  Kubernetes-based bare-metal provisioning. Accelerator platform bringup (NVIDIA, AMD, or custom). Burn-in and stress…