About this role
About Nscale
Nscale is the GPU cloud engineered for AI. We provide cost-effective, high-performance infrastructure for AI start-ups and large enterprise customers. Nscale enables AI-focused companies to achieve superior results by reducing the complexity of AI development. Our GPU cloud bolsters technical capabilities and directly supports strategic business outcomes, including cost management, rapid innovation, and environmental responsibility.
At Nscale, our Support and Operations team plays a critical role in maintaining service availability, driving service reliability and rapid response to customer tickets.
We thrive on a culture of relentless innovation, ownership, and accountability, where every team member takes pride in their work and drives it with excellence and urgency. As an Nscaler, you’ll build trust through openness and transparency, where everyone is inspired to do their best work. If you join our team, you’ll be contributing to building the technology that powers the future.
AI Reliability Engineer II - APAC
What You'll be Doing
• Join the Support duty rotation as a senior escalation point, collaborating with Infrastructure Engineering, CNPRE, Network Operations, and Product Engineering on incidents, investigations, and changes
• Diagnose and remediate GPU node faults across the full stack — driver, firmware, and hardware layers — from nvidia-smi/DCGM and XID/RAS analysis through BMC/Redfish and out-of-band management to physical fault isolation and vendor RMA
• Own east-west fabric health: run link-level diagnostics (mlxlink, ibdiagnet, or equivalent), isolate transceiver, optics, cabling, and switch-port faults, and validate topology across InfiniBand and RoCE/high-speed Ethernet fabrics.
• Investigate data-path issues on high-performance storage platforms (e.g. VAST), including storage–network interactions across clients, mounts, VIPs, and routing
• Run structured, hypothesis-driven investigations; conduct root cause analysis for major incidents and drive long-term fixes to completion
• Author and execute changes in live customer environments with proper risk assessment, peer review, and backout plans
• Proactively improve dashboards, alerts, and runbooks to prevent repeat incidents; identify recurring patterns and convert them into problem records and automation
• Accurately record, update, and resolve tickets, keeping internal and external parties informed with clear customer-impact statements and evidence-rich notes that enable clean handover
• Design and implement automation scripts and small tools to reduce toil and human intervention
• Act as a key escalation point for the Support Organisation; taking ownership of strategic decisions where results matter
• Mentor and upskill mid-level engineers; contribute to knowledge sharing across Operations and Engineering, including training content, workshops, and PR reviews
• Lead by earning trust and speaking candidly. Disagree when appropriate and challenge the status quo; commit wholly to decisions once in motion
• Respond to critical incidents out of business hours and participate in on-call as required. Travel to Nscale or customer sites to provide onsite technical expertise
About You
• 4+ years in infrastructure, operations, or support engineering in production environments; 2–3+ years hands-on with GPU, HPC, or large-scale data centre estates, ideally in a customer-facing or escalation-driven capacity
• Able to explain complex technical detail clearly, specifically, and concisely — in tickets, in incident updates, and face to face with customers and stakeholders at all levels. Strong written discipline: your notes let the next engineer pick up where you left off without starting from scratch
• GPU platforms (NVIDIA; AMD Instinct beneficial). Practical, current experience with GPU drivers, firmware, and runtime stacks on AI training and inference clusters. Confident with nvidia-smi, DCGM, and XID/error interpretation; able to isolate faults across GPU, baseboard, NIC, and PCIe layers and drive them through diagnosis to RMA
• High-performance east-west fabrics. Hands-on experience with RDMA fabrics such as InfiniBand and/or RoCE including link-layer diagnostics (mlxlink, ibdiagnet, or equivalent), transceiver and cabling fault isolation, and understanding of rail-optimised topologies, NVLink/NVSwitch, and NCCL-based performance troubleshooting on multi-node clusters
• HPC scheduling. Slurm operations for large multi-GPU jobs — containers via Pyxis/Enroot, MPI, and diagnosing queue, topology, and job failures
• Linux systems engineering at scale. Strong command of modern Linux distributions, kernel modules, systemd, networking stack, and filesystem tooling. Proven troubleshooting across compute, storage, and network layers in production
• Server hardware and control planes. Comfortable with BMC/Redfish, firmware management, and bare-metal provisioning workflows (MAAS or similar) across large node fleets
• Networking fundamentals. Solid grasp of L2/L3, routing, BGP, VLANs, VXLAN, firewalls, and load balancing, with a clear understanding of how east-west cluster traffic differs from north-south
• Observability and incident response. Build and use alerting stacks and dashboards (Prometheus/Grafana or similar), interpret metrics and alerts, drive runbooks to resolution, and contribute to SLOs and post-incident reviews
• Change and risk judgment. Experience authoring and executing changes in business-critical environments, including risk assessments, customer-impact analysis, and backout plans
• SRE-style operations. Write and maintain runbooks, automate diagnostics, and reduce human intervention through scripts and small tools
• Automation and Git. Scripting skills in Bash, Python, or equivalent for operational tooling and integrations; experience with infrastructure automation tools (Ansible, Terraform, or similar)
• Data Centre fundamentals. Understanding of how data centres operate — se
Description from Nscale's public careers feed, reproduced so you can read the role here. Apply on the company's own site; RealAnalystJobs never submits anything for you.
