Overview
We are looking for an experienced fleet infrastructure engineer to build the software systems that manage, orchestrate and maintain OneQode's growing AI/GPU fleet. As we grow, we need that fleet ready to be productised and sold as a service, built by someone who understands the physical infrastructure underneath, not a conventional DevOps engineer whose Kubernetes experience is only on public cloud. This is a senior, hands-on role, fully remote and based in APAC.
You will translate what our product team needs into a fleet that delivers it, much like a conductor turning a score into a performance. You will own fleet-wide provisioning, node lifecycle, orchestration and recovery, with the opportunity to revamp existing approaches and build your own.
Responsibilities
- Turn product requirements into infrastructure solutions, then package them and roll them out across data centres.
- Design and build fleet management systems for provisioning, configuration, monitoring, upgrades and automated recovery.
- Own node lifecycle, from bringing servers into service to draining, reimaging, upgrading and returning them to production.
- Run Kubernetes on bare metal, and integrate Kubernetes and Slurm with our provisioning and fleet management systems.
- Improve cluster reliability, availability and operational resilience across large-scale GPU fleets.
- Build and improve tooling for automated fault detection, node isolation, recovery and return to service.
- Implement and improve our infrastructure engineering standards and practices.
Preferred Skills and Experience
- 5+ years in infrastructure engineering, including significant production Kubernetes experience on bare metal.
- Strong Linux, networking, distributed systems and infrastructure automation skills.
- Experience building fleet management systems for provisioning, configuration, monitoring, upgrades and automated recovery.
- Strong software engineering skills, including building internal platforms and tooling.
- Understanding of physical server infrastructure, hardware failure modes and GPU cluster architecture.
- Ideally, experience with large-scale GPU clusters, Slurm, InfiniBand or RoCE/RDMA, and CUDA or ROCm.
- A background at a neocloud, hyperscaler, GPU cloud provider, HPC operator or large infrastructure platform team.
- Builder's mindset. You enjoy rethinking how things are done and building better systems.
- Product-minded. You measure success by what the fleet delivers for customers, not just by uptime.
We do not expect every candidate to have experience in all of these areas. Slurm experience is a plus, but if you have strong bare metal and fleet engineering fundamentals, good judgement and the right mindset, we would still like to hear from you.
About OneQode
OneQode is a global provider of performance digital infrastructure. Our vertically integrated platform spans cloud compute, low-latency networking and sovereign technology across more than 30 data centres on 5 continents. We are deploying and scaling AI infrastructure for customers, bringing together compute and connectivity to support their growth. We support enterprises, governments and performance-driven businesses in running AI and mission-critical workloads at scale, worldwide.