AI Kubernetes Platform Engineer
US client via staffing partner · Bay Area, CA
- Location
- Bay Area, CA · Onsite
- Salary band
- Band not stated
- Type
- Contract
- Level
- Senior
- Work authorization
- Not stated
Stack
Kubernetes · GPU workloads · NVIDIA GPU Operator · Linux administration · Helm · GitOps · Python · Bash · Go · etcd · CNI · CSI · Prometheus · Grafana · Loki · NVIDIA DCGM · Kubeflow · Argo · service mesh · infrastructure as code
About the role
Role: AI Kubernetes Platform Engineer Location: Bay Area, CA Contract Job Description: The engineer will build, administer, and troubleshoot Kubernetes platforms supporting model development, distributed training, inference, and shared AI services. The role requires deep cluster administration skills and practical experience with GPU workloads in bare-metal or data center environments, with responsibility for cluster lifecycle, tenant readiness, and platform reliability. Key responsibilities- - Provision and maintain Kubernetes clusters, including control plane services, worker nodes, container runtimes, configuration baselines, and cluster lifecycle automation. - Diagnose failures involving etcd, API services, kubelet, CNI, CSI, ingress, DNS, image pulls, scheduling, autoscaling, pod eviction, and degraded node states. - Configure and troubleshoot NVIDIA GPU Operator, GPU device plugins, driver and runtime dependencies, GPU resource discovery, and workload placement. - Implement tenant namespaces, quotas, role-based access controls, network policies, secrets integration, and appropriate workload isolation and GPU sharing policies. - Integrate storage and networking for AI workloads. Investigate throughput, mount, connectivity, and resource-contention problems with Linux, storage, and fabric teams. - Automate platform deployment and configuration with infrastructure as code, Helm, and GitOps. Plan upgrades with compatibility checks, validation tests, backups, and rollback procedures. - Build platform health checks and dashboards using metrics, logs, events, and GPU telemetry. - Validate clusters with representative workloads before production handover and after changes. - Maintain day-2 runbooks, capacity guidance, and escalation criteria. Work with SRE and validation teams on service objectives, resilience, and recurring failure prevention. Required experience and skills- - 7+ years in infrastructure engineering, including deep hands-on Kubernetes administration and cluster troubleshooting experience. - Strong knowledge of Kubernetes internals, control plane operation, node lifecycle, scheduling, container runtimes, and resource isolation. - Experience supporting GPU workloads on Kubernetes in production, laboratory, or validation environments, including device plugin and driver dependencies. - Strong Linux administration and practical understanding of network, DNS, storage, and authentication dependencies in data center environments. - Hands-on experience with Helm, GitOps or declarative operations, configuration management, and automated cluster provisioning or upgrades. - Python, Bash, or Go scripting skills and the ability to trace failures across node, pod, network, storage, and control plane layers. Preferred experience- - Experience with bare-metal Kubernetes, Kubeflow, Argo, service meshes, and high-performance storage integration. - Familiarity with Prometheus, Grafana, Loki, NVIDIA DCGM, GPU sharing, and Multi-Instance GPU where supported. - Experience with platform hardening, controlled production changes, disaster recovery exercises, and multi-tenant AI environments. Recommended certifications- - Preferred: Certified Kubernetes Administrator (CKA) and NVIDIA-Certified Professional AI Infrastructure (NCP-AII). Security specialization: Certified Kubernetes Security Specialist (CKS). Additional operations credential: NVIDIA-Certified Professional AI Operations (NCP-AIO). Foundational: NCA-AIIO. - Certifications are preferred credentials. Candidates should demonstrate the corresponding skills through practical delivery and troubleshooting experience. Expected deliverables and success measures- - Deliver reproducible cluster configurations, validated GPU enablement, documented upgrade and recovery procedures, and operational dashboards. Success is demonstrated by reliable workload placement, completed readiness checks, effective tenant isolation, and resolution of recurring platform faults.
Recruiter contact
Just $1 to see this recruiter's name, email and phone. One-time, per role — no profile needed.
Want the full job-search package instead? Talk to us →
The recruiter's details are not public. Helen makes the introduction, checks your resume against the requirements first, and follows up on the reply.