AI Reliability Engineering Manager

The role

Run the day-to-day operation of an AI Reliability Engineering team of eight to ten engineers across EMEA. You own the people, the queue, the rota, and the escalations. You report to the AIR Senior Manager for EMEA, who holds strategy, budget, and the senior customer relationships, so your time goes on running a world-class operation and growing the engineers in it.

The estate is about 29,000 GPUs today, with a planning target of one million by the end of 2027. You are the first manager hired under the Senior Manager, and as the region grows you take ownership of a full pillar with its own queue and SOPs. It is a hands-on role: close enough to GPUs, Linux, and high-performance fabrics to lead an incident and challenge a diagnosis, while leading the team most of the week.

What you'll be doing

• Lead a team of eight to ten AI Reliability Engineers: 1:1s, objectives, reviews, probation, development, and hiring
• Own the queue: prioritization, SLA adherence, MTTR, and backlog, with a named owner on every near-breach ticket and clean handoffs to Engineering and vendors
• Run the rota and on-call for the team within the EMEA coverage model and the global follow-the-sun handover
• Be the first management escalation point on P1s and customer escalations, lead post-incident reviews, and turn repeat incidents into problem records and durable fixes
• Own the SOPs for your pillar and drive them into Nscale's automation so that repeatable remediation needs less human time
• Work the day-to-day relationships with DC Operations, Network Operations, Engineering, and the fleet tooling and automation teams
• Report your team's numbers weekly and accurately, and explain what moved and why

About you

• Experience: 6+ years in infrastructure, operations, or support engineering in production environments, including 2+ years around GPU, HPC, or large data center estates
• Leadership: 3+ years managing engineers in an operational support function, owning reviews, development plans, and performance management through to outcome
• Operational ownership: you have run a queue against SLAs and are comfortable being accountable for the numbers
• Communication: clear, specific, and concise, in writing and in person
• Decisiveness and pragmatism: you make calls with incomplete information, improve process without slowing delivery, and commit once the decision is made
• Technical foundation: solid production Linux and networking, working knowledge of how GPU fleets fail and get repaired, ITIL and SRE habits, and enough scripting to review an engineer's work
• Adaptability: comfortable with out-of-hours escalations, on-call, and travel to data centers and customer sites

Nice to have

• NVIDIA GPU platform depth: nvidia-smi, DCGM, and XID interpretation through to RMA (AMD Instinct beneficial)
• InfiniBand or RoCE diagnostics, NCCL, NVLink and NVSwitch, and Slurm with Pyxis and Enroot
• Rack-scale systems such as GB200 and GB300 NVL72
• BMC and Redfish, MAAS, NetBox, OpenStack, Kubernetes, and Ansible or Terraform
• High-performance storage such as VAST, Ceph, or NFS at scale
• Follow-the-sun operations, and ITIL, Linux, networking, Kubernetes, or NVIDIA certifications