Senior Staff Engineer – ML Ops

About the role

Shield AI is building a Kubernetes‑native AI Factory Reference Architecture to support the development, training, evaluation, and deployment of advanced autonomy systems. As a Senior Staff Engineer you will shape the platform’s architecture and implementation while collaborating with ML researchers and autonomy teams.

Key responsibilities

• Lead design and implementation of the AI Factory Reference Architecture, delivering a Kubernetes‑native platform for AI development, distributed training, simulation, evaluation, and deployment.
• Partner with ML researchers to integrate state‑of‑the‑art frameworks, foundation model development, reinforcement learning, and emerging workflows.
• Design self‑service developer experiences that move engineers from local experiments to large‑scale distributed execution.
• Build and optimize shared GPU infrastructure across cloud and on‑premises environments, improving utilization, scheduling, storage, networking, and observability.
• Implement data and model lifecycle capabilities, including dataset management, experiment tracking, model versioning, and continuous improvement.
• Develop repeatable deployment and lifecycle management using Infrastructure as Code for commercial cloud, sovereign, and air‑gapped environments.

Required profile

• Proven experience building Kubernetes‑native AI or MLOps platforms for distributed machine‑learning workloads.
• Deep knowledge of modern AI training frameworks such as PyTorch and Hugging Face Transformers.
• Hands‑on experience operating GPU‑accelerated infrastructure and large‑scale distributed training systems.
• Strong understanding of Kubernetes, Linux, networking, security, storage, and distributed systems.
• Experience packaging cloud‑native infrastructure with Terraform and Helm.
• Solid software engineering skills in Python and Golang.

Required skills

• Kubernetes
• Python
• Golang
• PyTorch
• Hugging Face Transformers
• Terraform
• Helm
• Ray (preferred)
• Slurm / KAI (GPU scheduling)
• OpenTelemetry
• Prometheus
• Grafana