Customer Deployment Technician

About QumulusAI QumulusAI operates high-performance bare-metal GPU infrastructure at carrier-neutral exchange points, serving AI and ML customers (inference platforms, model developers and compute aggregators) who buy on performance, price and the credibility of our operations. We run our own network, our own fabrics across the latest Ethernet and InfiniBand, and our own control plane across multiple production data centers. We build at hyperspeed. Rather than wait years for a greenfield campus to come online, we take existing data center capacity that is ready today and put it to work, deploying our own servers, network stack, storage, racks and PDUs, and turning a site up in 90 days. The model is distributed by design, and the long-term goal is to operate in more than a hundred locations. About the role We are hiring Customer Deployment Technicians to take our GPU clusters from racked hardware to a cluster a customer can use. You are the independent check on every hardware build: you check what the install crew built, bring it to our firmware baseline and get every link up. Then you drive the build through our automation, clear what it flags, test the cluster the way the customer will, and hand it over secure and documented. This is a hands-on technical role. You run each build day to day, from racked hardware to handoff, and you are the one who proves a cluster is ready. Most bring-up problems are in the network, so expect to spend real time on optics and fiber as well as Linux. Hands-on here means technical, not physical: you check and test hardware remotely through BMC, switch port and optics diagnostics, and travel is for site visits and key milestones. The role is new and the automation is getting better every month. The more builds you run, the more of what you learn goes back into it, and the harder the problems you get to work on. We are on a hyper-growth trajectory and need to be ready. What you will do Independently check every hardware build and burn in every node, so no connection, firmware, link or early hardware fault reaches bring-up. Bring firmware, BIOS and drivers to our baseline, and keep them consistent across every node. Bring up and verify the network: check switch configuration and cabling against the plan, and confirm link health across InfiniBand and Ethernet. Drive each build through our automation, and clear what the dashboards flag. When something fails, find what is wrong and where, and escalate with the evidence. Direct remote hands and install crews at the site to fix cabling and hardware faults, with clear instructions and evidence. Prove every node and the cluster meet our baseline, tested as the customer will. Run performance benchmarks across the cluster, and resolve anything that falls short of expected results before handoff. Keep asset, configuration and test records accurate in NetBox, and log faulty parts with the evidence for RMA. Write and keep current the runbooks for each build step you own. Improve the automation with what each build teaches you, working with Engineering. Hand over a secure, documented cluster to Cloud Ops, and help Customer Success in the first days. Keep the Project Manager informed, and raise any failure the same day. Your first 90 days Days 1 to 30: learn our builds, validate your first racks with evidence on record, and run burn-in and node tests on a live build. Days 31 to 60: carry a build from hardware validation to handoff, take a second in parallel, and hand over your first cluster. Days 61 to 90: carry several builds at once, and take ownership of one more build step, with the method written down. What we are looking for Typically 3 to 5 years in Linux systems administration, server deployment or data center operations, or equivalent experience. Working knowledge of optics and fiber, including low-light and no-link faults. Working knowledge of Ethernet networking and switch configuration, enough to trace a fault from the server port to the switch. Comfortable at the command line and with scripts. Can run our Ansible playbooks and read them well enough to tell a playbook failure from a hardware fault. Has directed remote hands or field technicians to fix hardware faults. Troubleshoots from evidence, step by step, and can explain what was ruled out. Writes clear runbooks and test records that someone else can follow. Bonus points GPU servers and the NVIDIA software stack. If you don't have it yet, we'll teach you our GPU test tools and automation. InfiniBand or RoCE experience. BMC and Redfish experience. Has run cluster benchmarks such as NCCL or HPL. Hands-on with Dell, Supermicro or Lenovo servers and their BMC tools, such as iDRAC or IPMI. Tools you'll use Linux, Ansible, Redfish and BMC tools, NetBox, Zabbix, NVIDIA DCGM, NCCL and HPL, across InfiniBand and Ethernet fabrics. How you will know it is working Clusters are ready on the date in the plan. Every connection and every node has evidence on record before handoff. Customers find no issue in their first week that our tests should have caught. Every escalation comes with a likely cause, not just a symptom. Who thrives here You take it personally when something ships half-checked. A cluster is done when it is proven, not when it is plugged in. You would rather find the bad optic than guess at it, and you can say roughly what is wrong before you escalate. You are comfortable moving between server vendors, GPU platforms and network gear. No two of our clusters are the same. You like running several things in parallel and keeping every one of them on track. You notice when a step could be automated, and you say so. How we work with AI AI tools are part of how we work here, and everyone is expected to use them daily: to draft, to check their own work and to move faster. The tool helps; you own the result. Nothing goes to a customer without a person reading it and standing behind it. Travel and hours Remote, with travel up to 25% for site visits, build kickoffs and key milestones. Build and test weeks can fall outside normal business hours, and you will flex your hours when a build needs it. Why QumulusAI QumulusAI is building the next generation of AI infrastructure, and we are early enough that the decisions you make will still be how this company operates at a hundred sites. We offer competitive compensation, equity participation, and infrastructure problems at a scale few companies encounter. The base salary range for this role is $144,000 to $192,000, set by experience and skills, plus equity participation. QumulusAI is an equal opportunity employer. All qualified applicants will receive consideration without regard to any characteristic protected by law.