Senior Hardware Diagnostics Engineer - Infrastructure at Tensorwave in Las Vegas, NV

Job description

About TensorWave Our mission is simple: deliver seamless, secure, reliable, and resilient AI compute at scale. We've built a versatile cloud platform that eliminates infrastructure barriers, empowering builders to focus on innovation instead of fighting their stack. Because breakthrough AI should move at the speed of ideas, not infrastructure. About the Role We are building and operating large-scale infrastructure platforms to support high-performance AI workloads across multiple data centers. Our environment includes GPU-intensive systems, high-throughput networking, and rapidly scaling compute clusters. We are looking for a Senior Hardware Diagnostics Engineer to own that surface: which versions we run, how we prove a version is safe, and how we push it across a fleet without breaking production. You'll set the bar for what validated hardware means here, build the tooling that enforces it, and be the person we escalate to when a failure doesn't make sense. Firmware is where a lot of our hardest problems live. A BIOS revision changes how a GPU enumerates. A BMC update fixes one thing and breaks telemetry. What You’ll Do Own the firmware lifecycle across BIOS, BMC, NIC, GPU, and storage controllers: version qualification, approved baselines, staged rollout, and rollback Design and build the validation harness that qualifies a firmware release before it reaches production, including regression and performance comparison between versions Build automation against Redfish, IPMI, and vendor management APIs so firmware and validation work runs at fleet scale rather than node by node Define what passing burn-in means: the test matrix, the thresholds, and the evidence we keep Make fleet-wide changes safe and predictable through pre-flight validation, blast radius control, staged waves, and a rollback path that has actually been tested Lead root cause analysis on hardware and firmware failures nobody else can reproduce, including PCIe error analysis, sensor and event log forensics, and thermal and power telemetry Own technical escalation with OEM, ODM, and silicon vendors on firmware defects, and track fixes to closure Prepare validation and bring-up capability for next-generation AMD platforms before the hardware lands Mentor the diagnostics engineer on the team, and turn recurring findings into runbooks and tooling Participate in incident response and root cause analysis Contribute to runbooks, documentation, and operational best practices Who You Are Required Qualifications 6+ years in infrastructure engineering, hardware validation, systems engineering, or a closely related field Deep hands-on experience with BMCs and out-of-band management — Redfish, IPMI, iDRAC, iLO, or equivalent — including scripted interaction, not clicking through a web UI Direct firmware experience flashing, qualifying, and troubleshooting BIOS, BMC, and device firmware on enterprise servers, and a clear sense of how badly this can go wrong at scale Strong Linux systems skills: boot process, kernel and driver behavior, and hardware diagnostics such as {{dmesg}}, {{lspci}}, {{ipmitool}}, SMART, and vendor CLIs Ability to read low-level hardware signals: PCIe topology and error output, SEL and sensor data, thermal and power telemetry Python for automation against REST APIs — Redfish, vendor tooling, and internal systems A rigorous diagnostic method: you form a hypothesis, design a test that can disprove it, isolate one variable at a time, and document the evidence behind your conclusion Experience driving hardware issues with vendors through to resolution Clear technical writing for runbooks, RCAs, and vendor escalations Preferred Qualifications AMD GPU and ROCm experience: {{amd-smi}}, {{rocm-smi}}, ROCm Validation Suite GPU and HPC validation at scale: stress testing, collective communication testing with RCCL or NCCL, node health frameworks Low-level bus and firmware work: I2C, SMBus, BMC firmware behavior, Redfish schema extensions Kernel-level debugging,