Principal AI/ML Platform Engineer - Remote

About the role

Optum Tech is a global leader in health care innovation. Our teams develop cutting-edge solutions that help people live healthier lives and help make the health system work better for everyone. From advanced data analytics and AI to cybersecurity, we use innovative approaches to solve some of health care's most complex challenges. Your contributions here have the potential to change lives. Ready to build the next breakthrough? Join us to start Caring. Connecting. Growing together.

As a Principal AI/ML Platform Engineer on the UnitedHealth Group (UHG) enterprise team, you will serve as the AI Architect across IaaS and PaaS environments, owning the technical direction, reference architecture, and long-term evolution of our multi-tenant AI compute platform end to end. Our team builds and maintains an advanced compute estate spanning on-premises bare-metal Red Hat OpenShift AI clusters equipped with high-performance NVIDIA GPUs and InfiniBand/RoCE training fabrics, alongside public-cloud managed AI services including Azure AI Foundry, AWS Bedrock, and GCP Vertex AI.

Responsibilities

• Own the end-to-end reference architecture for multi-tenant AI compute platforms across hybrid on-premises bare-metal OpenShift AI clusters and public-cloud managed AI platforms (Azure AI Foundry, AWS Bedrock, GCP Vertex AI)
• Set network, latency, and topology standards for distributed training (NVLink, InfiniBand, RoCEv2, GPUDirect RDMA, NCCL), ensuring interconnect boundaries are strictly maintained
• Establish cluster governance, GitOps workflows (Argo CD), RHACM policies, and automated lifecycle management for bare-metal accelerated compute nodes
• Design cost and utilization models including capex amortization, accelerator-sharing strategies (MIG/time-slicing), and cost-per-token/training-run math to inform accelerator procurement roadmaps
• Standardize model-serving platforms (vLLM, KServe) and inference gateways, setting quantization policies, provenance review gates, and intelligent model routing rules
• Define workload placement frameworks to determine self-hosted versus managed cloud deployment based on data residency, latency, cost, and compliance requirements
• Design and enforce identity, access, and security controls for AI workloads and autonomous agents, including least-privilege RBAC, Vault secret management, short-lived credentials, and mTLS
• Establish enterprise Service Level Objectives (SLOs), disaster recovery plans, and upgrade strategies