Key Responsibilities

Physical Infrastructure &

- Layer 1 Execution

- Rack &
- Stack Execution:

- Physically install and verify DGX/HGX/MGX nodes, switches, and PDUs against the current rack elevation and LLD.
- Confirm floor-loading, bolting, and levelling before energisation; escalate any structural discrepancy to the Domain Architect AI Facilities.
- Structured Cabling &
- P2P Validation:
- Execute the point-to-point (P2P) cabling schedule; confirm transceiver type and MPO cable size against the code on the box, not the colour, before patching.
- Clean and inspect optical connectors on every patch; validate seating and troubleshoot link-down, miswire, and link-flap faults by elimination (reseat, swap to a known-good port, clean, replace).
- OOB Standup:
- Bring up and validate the out-of-band (BMC/IPMI) management network ahead of in-band and compute-fabric activation, keeping it physically segregated per design.
- Confirm node power state and basic health via BMC before handing off to HPC configuration. 2. HPC Compute, Network &
- Storage Configuration
- Bare-Metal &
- Fabric Provisioning:
- Execute NVIDIA Base Command Manager (BCM) provisioning workflows and Ansible playbooks supplied by the Domain Architects to bring compute nodes, switches, and storage clients into service.
- Configure host-side networking (IPoIB, Netplan) and mount high-performance storage clients (VAST, WEKA, Lustre) to the current LLD.
- Firmware &
- Driver Lifecycle:
- Execute SBIOS, BMC, GPU VBIOS, and NVSwitch firmware upgrades per the NVIDIA firmware recipe across compute, network, and storage tiers.
- Apply OS hardening, kernel patching, and driver updates during scheduled maintenance windows.
- Performance Validation:
- Execute and log HPL, NCCL-tests, ib_write_bw/ib_send_bw, and IOR/FIO benchmark suites; compare results against the Gold Standard and flag deviations to the relevant Domain Architect. 3. RMA/DOA Diagnosis &
- Operations Support
- Fault Isolation:
- Triage Dead-on-Arrival hardware,



Xid errors, flapping links, and stale storage mounts, isolating whether the fault sits in compute, network, or storage before escalating.
- Distinguish a “Node Issue” from a “Network Issue” or a “Storage Issue” using nvidia-smi/dmesg, ibstat/ibdiagnet, and iostat/iotop.
- RMA Coordination:
- Raise and track Return Merchandise Authorisation cases with OEM and vendor support; coordinate physical swap-out with minimal schedule impact.
- L2 Ticket Resolution:
- Handle on-site escalations passed from L1 support; close the loop with the offshore HPC Engineer squad on root cause and remediation.

Technical

CompetenciesEssential Skills

- Data Centre Background: Prior Data Centre Technician, Field Engineer, or rack-and-stack experience within a System Integrator, OEM, or colocation setting.
- Fault Diagnosis:
- Comfortable triaging GPU faults (nvidia-smi, dmesg), link faults (ibstat, ibdiagnet, ethtool), and storage mount issues (iostat, iotop) to isolate a fault domain before escalating.
- Physical Infrastructure (Layer 0/1):
- Rack-and-stack procedures, structured cabling (OS2/OM4/DAC/AOC), MPO and OSFP transceiver handling, and cable pathway standards.
- Site Acceptance Test (SAT) support and As-Built documentation capture. Desirable Experience
- Linux &

- Automation:

- Solid RHEL/Ubuntu administration; ability to execute and troubleshoot Ansible playbooks.
- Git workflow familiarity (pulling code, branching, committing configuration changes).
- HPC &
- AI Infrastructure:
- Working proficiency with NVIDIA Base Command Manager (BCM) for bare-metal provisioning.
- Familiarity with DGX/HGX/MGX hardware architecture and standard benchmark suites (HPL, NCCL-tests, IOR/FIO).
- Network Affinity: Hands-on exposure to InfiniBand/RoCEv2 cabling and switch-side transceiver handling.
- Storage Integration: Parallel filesystem client mounting experience (VAST, WEKA, Lustre).
- Container Platforms: Kubernetes/Red Hat OpenShift node-level troubleshooting (NVIDIA GPU Operator, Network Operator).

📌 Field Solution Engineer (Sydney)
🏢 Nityo Infotech
📍 Sydney