Principal AI/ML Engineer Remote Nationwide or Office-Based in MN/DC/NJ

UnitedHealth Group•Eden Prairie, MN

9d•$164,600 - $282,200•Remote

About The Position

The Principal AI/ML Engineer for GTS is a director-level, hands-on individual contributor who researches, designs, builds, and ships enterprise AI solutions for healthcare applications and workflows. The role has enterprise-wide influence and is directly accountable for production code and delivery, from early experimentation through secure, reliable operation. The engineer partners frequently with CIO-level executives and senior business and technology leaders, translating strategic priorities and ambiguous requirements into research questions, working demonstrations, and production-ready solutions. They communicate evidence, limitations, risks, and architectural trade-offs, and recommend whether to scale, refine, or stop an approach. The technical scope includes generative AI, LLMs, agentic systems, RAG, and applied machine learning. LLM-assisted coding is a core practice, with independent validation and rigorous standards for correctness, security, maintainability, and responsible AI. The role is remote anywhere in the U.S. or office-based in Minnesota, Washington, D.C., or New Jersey. Hires in the Minneapolis or Washington, D.C. area must work in the office at least four days per week. Remote employees must follow UnitedHealth Group’s Telecommuter Policy.

Requirements

Bachelor’s degree in computer science, engineering, mathematics, or a related field, or equivalent practical experience demonstrated through engineering delivery.
12+ years of software engineering experience.
8+ years writing, testing, and maintaining production application or platform code.
3+ years designing, deploying, and operating production AI/ML capabilities, including model or service integration, evaluation, release controls, monitoring, and ongoing improvement.
3+ years deploying and operating services on at least one cloud platform, including CI/CD, logs, metrics, traces, incident diagnosis, and release or rollback procedures; experience implementing access controls, secrets management, and safeguards for sensitive data.
2+ years designing and shipping LLM-powered applications or agentic systems, including RAG and operational agent workflows with tool integration, multi-step orchestration, state management, permission controls, failure recovery, and human approval where needed.
2+ years using LLM-assisted coding tools to design, implement, and iterate on complex software systems.

Nice To Haves

Experience delivering healthcare applications or AI solutions involving protected health information or other regulated data, including privacy controls, auditability, and responsible AI evaluations.
Experience partnering directly with CIO-level executives on technology research, solution demonstrations, investment trade-offs, and production adoption.
Experience creating reusable AI platforms, agent services, reference implementations, or developer tools adopted by multiple engineering teams.
Experience with multi-agent orchestration, agent memory, Model Context Protocol (MCP), or equivalent enterprise tool-integration approaches, including testing and monitoring multi-step failure modes.
Experience with model adaptation or fine-tuning, open-source model serving, inference optimization, or advanced retrieval and ranking techniques, supported by comparative evaluations.
Experience with containers, Kubernetes, infrastructure as code, model registries, and automated evaluation pipelines supporting repeatable software and model releases.

Responsibilities

Partner with CIO-level stakeholders to identify opportunities, clarify constraints, shape technical priorities, and present solutions, findings, risks, and recommendations.
Lead time-boxed research and experiments evaluating models, agent architectures, frameworks, and AI development tools; establish baselines and success criteria and make evidence-based build, buy, or reuse decisions.
Personally design, implement, test, deploy, and operate production AI services and applications, owning the transition from prototype through enterprise integration and ongoing support.
Build bounded agentic workflows that plan, retrieve context, use tools, and perform multi-step tasks; implement orchestration, state and memory management, structured outputs, retries, timeouts, recovery, and human approval where appropriate.
Use LLM-assisted coding to explore architectures, implement features, generate and refine tests, debug, and refactor; independently validate code, dependencies, security, and behavior.
Develop RAG and semantic-search pipelines with appropriate ingestion, retrieval, ranking, and access controls; select models, prompts, and tools based on demonstrated performance and use conventional ML or simpler software when appropriate.
Create representative test datasets, automated evaluations, and regression suites for task completion, tool-call correctness, groundedness, safety, latency, and cost; use telemetry and feedback for continuous improvement.
Implement security, privacy, and responsible-AI measures, including least-privilege access, secrets management, safe data handling, audit trails, execution isolation, guardrails, fallback paths, and human oversight.
Lead cross-team architecture decisions and develop reusable services, reference implementations, and engineering standards; mentor senior engineers through code and design reviews.
Deliver demonstrable releases at pace, adapt to evidence and changing priorities, surface blockers, and track results against business and engineering measures.