Technical Lead - GPU Infrastructure

About the role

This is a hands‑on technical leadership role responsible for architecting and delivering a large‑scale GPU infrastructure platform. You will lead the transition from managed Kubernetes workloads to bare‑metal GPU infrastructure, supporting research, model‑training and managed inference workloads.

Key responsibilities

• Own end‑to‑end platform architecture, including proposals, designs, technical reviews and documentation.
• Lead and line‑manage a distributed engineering team across backend, frontend, DevOps, QA and documentation.
• Design, build and operate a managed Slurm service for research and model‑training workloads.
• Lead GPU infrastructure operations on bare‑metal, handling NVIDIA drivers, CUDA, Fabric Manager, NVSwitch, DCGM and MIG.
• Bootstrap and maintain Kubernetes clusters on bare metal using NVIDIA GPU Operator and Network Operator.
• Define managed inference architecture covering serving, multi‑GPU parallelism, autoscaling and confidential‑compute capabilities.
• Establish observability across control plane, GPU fleet and applications with metrics, logging, alerting and SLOs.
• Lead incident response, on‑call model design and post‑incident reviews.

Required profile

• 8+ years of hands‑on engineering experience, with at least 3 years leading infrastructure platform teams.
• Bachelor’s or Master’s degree in Computer Science, Engineering or related field, or equivalent practical experience.
• Extensive production experience operating Slurm, HPC or GPU training clusters and NVIDIA GPU fleets on bare metal.

Required skills

• Slurm (slurmctld, slurmdbd, partitions, QoS, accounting, upgrades).
• Kubernetes (control plane, operators, multi‑tenancy, GPU Operator).
• NVIDIA driver and CUDA lifecycle, Fabric Manager, NVSwitch, DCGM, MIG.
• InfiniBand, subnet configuration, RDMA, SR‑IOV, NCCL performance tuning.
• Linux kernel modules, PCIe passthrough, vfio‑pci, cgroups, namespaces.
• Observability tools: Prometheus, Grafana, Loki.
• JavaScript and Node.js for reviewing control‑plane and worker services.
• Experience with HPC storage (Lustre, VAST, NFS) and NVMe caching.

What we offer

• 100 % remote work with flexibility across UTC to UTC+5:30 time zones.
• Leadership of a sophisticated GPU infrastructure platform spanning bare‑metal, Slurm, Kubernetes and inference.
• Opportunity to shape architecture for advanced AI research, model training and managed inference workloads.