Principal Data Scientist

Optum is a global organization that delivers care, aided by technology to help millions of people live healthier lives. The work you do with our team will directly improve health outcomes by connecting people with the care, pharmacy benefits, data and resources they need to feel their best. Here, you will find a culture guided by inclusion, talented peers, comprehensive benefits and career development opportunities. Come make an impact on the communities we serve as you help us advance health optimization on a global scale. Join us to start Caring. Connecting. Growing together.

As a Senior Data Scientist (SG 29), you will hands on design, build, and productionize Agentic AI and GenAI solutions on top of a Healthcare Core Data Platform. You will deliver reliable, compliant, and scalable AI systems that work across large healthcare datasets (claims/clinical/provider/member) and enable measurable improvements in quality, cost, and operational efficiency.

Primary Responsibilities:

• Agentic AI & GenAI Delivery
  • Design and implement agentic AI systems (multi step, tool using agents) that can plan, execute, and verify outcomes under defined guardrails for healthcare workflows
  • Build GenAI solutions using enterprise approved LLMs including Claude 4.6 and OpenAI Codex (and equivalents) for:
    • intelligent data exploration and analytics assistance
    • automated insight generation and summarization
    • engineering productivity accelerators (code generation, refactoring, test creation)
    • workflow automation and triage support
  • Develop hybrid systems combining LLMs with classical ML (predictive/prescriptive models) for robust performance on healthcare use cases
• Core Data Platform Integration
  • Implement AI solutions tightly integrated with the Core Data Platform (curated datasets, standardized semantics, governed access, reusable components)
  • Partner with data engineering/platform teams to implement scalable patterns for:
    • secure tool-calling and controlled data access
    • retrieval / grounding patterns (enterprise search, knowledge bases, curated datasets)
    • reusable agent skills and shared libraries
• Model / Agent Lifecycle Ownership
  • Own end to end delivery: problem framing - data readiness - prototyping - evaluation - production deployment - monitoring and iteration
  • Build and maintain evaluation frameworks for LLM and agent behavior (quality, hallucination risk, safety, latency, and cost)
  • Implement drift and behavior monitoring for models and agents; create feedback loops and runbooks to maintain performance over time
  • Ensure production readiness: reliability, observability, incident response, and cost controls
• Governance, Security & Responsible AI

• Build solutions compliant with healthcare privacy and governance expectations (e.g., PHI handling, access controls, auditability, retention, and policy adherence)
• Implement guardrails: prompt protections, tool use restrictions, sensitive data redaction, and explainability approaches where required
• Document system behavior, limitations, and risk mitigations for technical and non technical stakeholders

• Bachelor's or Master's degree in Computer Science, Data Science, Statistics, Engineering, or related field
• 12+ years of overall experience, with 5+ years in Data Science / AI/ML / Applied AI in enterprise environments
• Hands-on experience with ML frameworks: Scikit Learn, and at least one of TensorFlow / PyTorch
• Experience building evaluation + monitoring for ML/AI systems (metrics, drift, observability, reliability)
• Practical experience applying LLMs and coding assistants in real workflows, including Claude 4.6 and OpenAI Codex (or equivalent enterprise-approved tools/models)
• Solid proficiency in Python (pandas, numpy) and SQL; solid debugging and problem-solving skills
• Proven track record delivering production-grade AI/ML solutions end to end (not just experimentation)

• Hands-on experience with agentic orchestration patterns: tool calling, memory strategies, guardrails, and multi-step workflow design
• Experience with streaming/event-driven systems (Kafka) for near-real-time use cases
• Exposure to LLMOps / AgentOps: prompt/version management, automated evaluation, red-teaming, telemetry, CI/CD integration
• Familiarity with big data / distributed compute (PySpark, distributed SQL engines) and large-scale pipelines
• Healthcare domain familiarity: claims/clinical/provider/member data and regulated delivery environments (privacy/security/compliance)
• Proven ability to own complex problem spaces end-to-end with minimal supervision, from design through production operations
• Proven ability to drive impact primarily through technical execution and reusable artifacts, not people management
• Proven ability to make engineering tradeoffs across accuracy, safety, latency, reliability, compliance, and cost and documents decisions clearly
• Proven ability to produce solutions that are adoptable and repeatable across multiple assets via templates, libraries, and reference implementations

• Production Adoption: agent/GenAI capability used by multiple downstream teams or workflows
• Quality & Safety: measurable improvements in evaluation scores; reduced hallucination/safety incidents via guardrails
• Operational Excellence: monitoring coverage, drift detection, and incident response readiness implemented
• Productivity: reduced cycle time for analytics/engineering workflows via Claude 4.6/Codex-enabled accelerators
• Governance: audit-ready documentation, controlled PHI access, compliant deployments