Staff, Platform Engineer AI Compute (6034)

Shield AI•San Mateo, CA

About The Position

Shield AI is seeking a hands-on software and platform engineer to improve the reliability, usability, and efficiency of the AI Factory, a shared compute and development platform for machine-learning engineers and customer deployments. The role focuses on software and configuration that enable developers to run GPU workloads on Kubernetes, predictable sharing of compute, workload troubleshooting, and using performance data to guide capacity planning. The engineer will help build the reference architecture used by Shield AI engineering teams and delivered to customers. The work will affect researchers’ ability to iterate, compute utilization, and the path from experimentation to operational deployment, across cloud, on-premises, and air-gapped environments.

Requirements

Experience building and supporting production Kubernetes platforms, including GPU or distributed computing workloads.
Strong software engineering skills in Python, Go, or a comparable language.
Experience automating infrastructure deployment, configuration, and upgrades.
Practical Linux, networking, and storage troubleshooting skills.
Experience with shared resource scheduling, access controls, or workload isolation.
Ability to work directly with machine-learning engineers, understand their workloads, and turn recurring problems into maintained platform capabilities.

Nice To Haves

Experience with Run:ai or comparable GPU scheduling and resource-management systems.
Familiarity with distributed training and tools such as Ray or PyTorch.
Experience delivering on-premises or air-gapped platforms, or contributing to open-source infrastructure.

Responsibilities

Build and maintain software and configuration that let developers run GPU workloads on Kubernetes.
Design quotas, scheduling priorities, permissions, and isolation so multiple teams can share compute predictably.
Automate deployment and upgrades of GPU services, including Run:ai, using infrastructure as code and tools such as Terraform and Helm.
Investigate why jobs wait, fail, or run slowly, and resolve recurring scheduling, configuration, storage, and networking problems.
Build diagnostics connecting queue times, GPU utilization, memory requirements, and training throughput to the developer experience.
Benchmark representative workloads and recommend when to add GPU capacity, how much to add, and which hardware fits the workload.
Deliver repeatable installations and upgrades across cloud, on-premises, and air-gapped environments.
Work directly with machine-learning engineers to understand their workloads and turn recurring problems into maintained platform capabilities.