Senior Fabric & Interconnect Reliability Engineer

Firmus Technologies; Sydney; On-site; AUD 140,000 - 210,000; Full time

Job summary

Firmus Technologies in Australia is seeking a Senior Platform Reliability Engineer (Fabric and Interconnect) to own the reliability of GPU interconnect fabrics, NVLink/NVSwitch, InfiniBand and related networks across a large AI infrastructure.

This hands-on senior role emphasizes automation, software-defined remediation, and collaboration with vendor engineers to drive permanent fixes, runbooks, and resilient operations.

Qualifications

• Extensive experience in high‑performance networking and production networks.
• Deep knowledge of GPU interconnect fabrics (NVLink/NVSwitch) and topology.
• Experience with InfiniBand, RoCE and related congestion control.
• Experience with DPU/SmartNIC based host networking and firmware coordination.
• Automation and infra‑as‑code practices in production environments.

Responsibilities

• Ensure reliable operation and continuous improvement of GPU interconnect and network fabrics.
• Develop guarded automation and remediation tooling for fabric faults.
• Diagnose performance across the interconnect stack and tune networks.
• Manage 24/7 incident response and on‑call escalation.
• Lead vendor escalations and drive durable fixes and runbooks.

Skills

High-performance networking

Production network ownership

24/7 operations experience

Incident response

Automation and SRE practices

Scripting (Python/Go/Bash)

Vendor escalation

Technical communication

Tools

NVIDIA NVLink

NVSwitch

InfiniBand

RoCE

Spectrum-X

DPU/SmartNIC

Job description

Firmus Technologies in Australia is seeking a Senior Platform Reliability Engineer (Fabric and Interconnect) to own the reliability of GPU interconnect fabrics, NVLink/NVSwitch, InfiniBand and related networks across a large AI infrastructure.

This hands-on senior role emphasizes automation, software-defined remediation, and collaboration with vendor engineers to drive permanent fixes, runbooks, and resilient operations.