Sr AI/ML Engineer - Remote Nationwide or Hybrid in MN/DC

Optum / UnitedHealth Group
Eden Prairie, MN / Remote Nationwide / Hybrid in MN/DC

About the job

Come build the AI foundation behind SGS, a modern platform where multi-agent systems can reason, plan, and safely connect to enterprise tools to turn conversations, operational signals, and network patterns into action. In this role, you will own the architecture for the capabilities customers feel every day, including real-time voice translation, voice-based assessments, and call quality audits that help teams respond faster and follow the right procedures. You will also advance our Virtual Investigator to accelerate intake, surface risk early, and turn complex investigative questions into clear, explainable answers. On the fraud side, you will combine real-time ML, graph analytics, and explainable AI to detect emerging fraud, waste, and abuse across claims, enrollment, provider, and encounter data; uncover collusive networks and billing anomalies; and translate those signals into defensible investigative leads, preventive controls, and KPI visibility leaders can use immediately. You will set the technical direction for orchestration and governance on Azure, raise engineering standards across product lines, and mentor a globally distributed team that ships to production. If you want deep ownership, high-stakes impact, and the chance to define how agentic AI operates at scale in a highly regulated healthcare environment, you will thrive here.

Responsibilities

AI Engineering Projects: Design and implement multi-agent AI systems that use LLMs, memory, and tools to reason, plan, and act autonomously

Function Calling & Orchestration: Build agent-based solutions that use function calling, dynamic tool integration, and orchestration frameworks such as LangChain, AutoGen, and Semantic Kernel

Modular Agent Design: Leverage standards such as Model Context Protocol (MCP) to define reusable, secure, and composable tool interfaces

Voice-Driven Interfaces: Develop voice-first AI agents using ASR technologies such as Whisper and Azure Speech, multi-turn conversation orchestration, and high-quality TTS

RAG & Memory Pipelines: Design and maintain retrieval and memory pipelines using vector databases and Azure Cognitive Search to ground agents in enterprise knowledge, prior interactions, and operational context

Fraud, Waste & Abuse Analytics & ML Ops: Design, build, and operationalize supervised and unsupervised models, including classification, clustering, anomaly detection, risk scoring, and graph/network analysis, to detect known and emerging FWA patterns across claims, enrollment, provider, and encounter data. Translate fraud typologies such as upcoding, unbundling, excessive units, duplicate or phantom billing, kickbacks, and encounter discrepancies into scalable model logic, rules, and real-time detection pipelines. Continuously refine detection effectiveness using referral, audit, and recovery outcomes where available

SQL & Data Engineering for FWA: Develop and optimize complex SQL queries, feature pipelines, and data validation checks for large-scale healthcare analytical workflows, including joins, window functions, aggregations, and performance-aware query design

Cloud-Native AI Deployment: Build, deploy, and monitor scalable AI services on Azure, including Azure OpenAI, Functions, Service Bus, Cosmos DB, Cognitive Search, and related tools

Explainability, Investigations & Governance: Produce clear, reproducible model outputs, narratives, visualizations, and KPI reporting that support investigators, clinicians, compliance teams, and business leaders. Contribute to model governance, validation, and documentation practices that ensure transparency, fairness, and regulatory defensibility

Agentic UX & AI as an Interface: Drive innovation in agentic user experiences, enabling AI to operate external tools and services securely on behalf of users

Mentorship & Collaboration: Review PRs, mentor junior engineers, and collaborate across India and US time zones in a distributed, agile environment

Qualifications

Minimum

Undergraduate degree or equivalent experience

5+ years of total engineering experience

5+ years of experience in AI/ML product engineering roles

5+ years of solid Python development experience; proficiency with ML frameworks such as PyTorch, scikit-learn, and Hugging Face

5+ years of experience with the Azure AI stack, including Azure OpenAI, Cognitive Services, Functions, Service Bus, and Cognitive Search

5+ years of experience with fraud detection, anomaly detection, risk scoring, or graph/network analytics pipelines

3+ years of solid experience with voice systems, including ASR, TTS, and real-time audio or telephony integration

2+ years of proven experience building and shipping LLM-powered or autonomous agent systems in production

2+ years of deep experience with LLM integration, tool calling, prompt engineering, and context-aware task execution

2+ years of hands-on experience with retrieval techniques such as RAG, semantic search, embeddings, and vector databases

Proven solid SQL development skills, including complex joins, window functions, aggregations, and performance optimization for analytical workloads

Experience working with healthcare claims, provider, enrollment, encounter, or other highly regulated transactional healthcare datasets

Demonstrated ability to explain model behavior, risk signals, and analytic findings to nontechnical stakeholders through clear, defensible documentation

Demonstrated track record of contributing to robust, testable, and scalable engineering systems

Preferred

Experience building automated evaluation harnesses for LLM and agent workflows, including golden datasets, offline and online testing, and measurable quality metrics such as task success rate, groundedness, or human-review agreement

Hands-on experience with responsible AI, including prompt injection testing, data exfiltration testing, safety reviews, and guardrails to reduce hallucinations and unsafe outputs in production

Proven experience implementing end-to-end observability for agentic systems, including distributed tracing, tool-call success rates, latency and error budgets, and token and cost telemetry with actionable alerting

Proven experience designing secure patterns for tool-enabled agents, including least-privilege access, secrets management, and policy-based controls for tool or API execution such as OAuth scopes, managed identity, and audit logging

Proven ability to optimize LLM or voice-system performance and cost using techniques such as caching, batching, streaming responses, rate limiting, model routing, and fallback strategies

Direct experience supporting Medicaid program integrity, healthcare fraud analytics, State Medicaid Agency analytics, or MCO SIU workflows

Familiarity with Medicaid reimbursement and billing constructs, including ICD-10, CPT/HCPCS, DRGs, revenue codes, NDCs, and encounter data

Familiarity with MMIS, T-MSIS, PERM, CMS program integrity guidance, or similar state or federal compliance frameworks

Experience supporting referral, recovery, audit, appeal, or case-prioritization workflows

Experience developing AI/ML solutions in regulated or government healthcare environments