Lead Systems Software Engineer – GPU & AI Infrastructure
Summary
Lead systems software engineer who architects and scales the software platform for SVI's on-site GPU data center: multi-node clustering with Kubernetes/Slurm, CUDA/NCCL and InfiniBand tuning, bare-metal automation via Ansible/Terraform, and Prometheus/Grafana telemetry for AI training and inference workloads.
About the role
SVI is seeking a Lead Systems Software Engineer (GPU Infrastructure) to architect, deploy, and scale the software platform powering our GPU data Center cluster. In this on-site role, you will bridge the gap between bare-metal accelerated hardware and the software stack required for high-performance AI training, inference, and compute workloads. Your primary responsibility will be creating a highly resilient, automated, and observable software layer that maximizes GPU utilization, eliminates networking bottlenecks, and delivers seamless scheduling capabilities for engineering teams.
Key responsibilities
- Design, implement, and maintain multi-node GPU clustering software using orchestration frameworks (e.g., Kubernetes with Volcano/Kueue or Slurm) to manage heavy batch processing workloads.
- Configure and optimize low-level drivers, runtimes, and libraries (NVIDIA CUDA, NCCL, Triton Inference Server) to ensure low-latency communication across the cluster.
- Identify and eliminate systemic performance drops across host CPUs, PCIe buses, high-speed interconnects (InfiniBand/RoCE), and high-performance storage systems (Lustre, GPFS, Weka).
- Develop automation scripts and configuration tools using Ansible, Terraform, and Helm to dynamically provision, patch, and deploy bare-metal server fleets.
- Build and manage specialized telemetry systems (Prometheus, Grafana, DCIM) to track GPU thermal metrics, power consumption, memory errors (XIDs), and utilization rates.
- Serve as the technical link between facility engineering (power/liquid cooling management) and product engineering teams running AI models.
About you
- 5+ years in Systems Engineering, Infrastructure, or DevOps, with at least 2 years directly managing production-scale GPU or HPC workloads.
- Proficiency in Python, Go, or C++ for system tooling and infrastructure automation.
- Advanced knowledge of Linux (Ubuntu/RedHat), networking internals (TCP/IP), and cluster networking like InfiniBand or RoCE (RDMA).
- Hands-on experience with production Kubernetes environments, Container Runtime Interfaces (CRI), and NVIDIA Container Toolkit integrations.
- Familiarity with data center GPU architectures (e.g., NVIDIA H100/B200), high-bandwidth memory (HBM), and NVLink interconnects.
- Experience with liquid-cooling telemetry or facility monitoring integrations (preferred).
- Contributions to open-source distributed training or batch scheduling projects (preferred).
- Certified Kubernetes Administrator (CKA) or formal NVIDIA/HPC certifications (preferred).
About us
SVI Public Company Limited is based in Pathum Thani, Thailand.
