Agentic AI Engineer, Healthcare AI

About the Role

Deloitte is building an AI-first effort to rebuild how the American healthcare system decides, aiming for faster, fairer, and less wasteful care. This is a ground-up rebuild of healthcare decision-making machinery at national scale. As an Agentic AI Engineer, you will own agent systems end-to-end, from architecture through production, with work shipping into live settings within months.

Key Responsibilities

• Agent Architecture & Orchestration
• Design and implement agentic systems for multi-step reasoning, planning, tool use, and workflow execution against complex, regulated operational processes.
• Build stateful workflows using frameworks like LangGraph and LangChain, including branching, retries, self-correction, human-in-the-loop checkpoints, and reusable orchestration patterns.
• Engineer for long-horizon reliability, including multi-step task completion, recovery from compounding errors, planning under uncertainty, and robust tool use.
• Build reasoning for regulated decisions, focusing on policy- and criteria-grounded outputs, proposer/critic/judge-style review, and auditable rationales for high-stakes decisions.
• Retrieval, Grounding & Context Engineering
• Develop end-to-end Retrieval-Augmented Generation (RAG) pipelines, covering ingestion, chunking, embeddings, vector and hybrid retrieval, reranking, contextual compression, and grounding strategies.
• Engineer memory and context management, including conversational state, persistent memory, retrieval-aware context assembly, and token-efficient context selection.
• Apply modern context-delivery patterns for agents to access the right information at the right time.
• Reliability, Evaluation & Safety
• Implement observability and tracing for prompts, tool calls, retrieval quality, agent traces, failures, drift, latency, and production behavior.
• Apply guardrails, safety controls, and failure-handling to reduce hallucinations and unsafe actions.
• Evaluate agents at the trajectory and task level, including multi-step task success, failure-mode and regression analysis, sandboxed test environments, retrieval- and generation-quality metrics, automated checks, and human review.
• Engineer healthcare-grade safety, including deployment eval gates, human-oversight and escalation models, auditability and traceability for regulated decisions, and PHI/HIPAA-aware data handling.
• Integration & Production Craft
• Build integrations with internal and external tools, APIs, enterprise systems, databases, and model providers.
• Deliver production-quality code with strong practices in testing, CI/CD, logging, versioning, and documentation; make architecture decisions balancing quality, safety, latency, cost, and model risk.
• Partner with modeling and post-training engineers to improve model behavior through evaluation-driven feedback and fine-tuned models.
• Translate complex operational processes into robust system logic and reusable AI patterns; stay current with advances in agentic systems and translate research into practical engineering decisions.

Requirements & Tech Stack

• LangGraph/LangChain or equivalent for stateful workflow building.
• Retrieval-Augmented Generation (RAG) pipeline development.
• Python engineering skills.
• Modern software practices: testing, CI/CD, version control, API integration.
• Experience implementing observability and tracing for LLM-based systems.
• Hands-on experience with at least one frontier model platform (Anthropic, Google, OpenAI) and/or open-weight/self-hosted models (e.g., Llama via vLLM).
• Ability to travel 0-50%.

Qualifications

• Bachelor's degree in Computer Science, Engineering, Data Science, Computational Linguistics, or a related field.
• Demonstrated depth building and shipping production agentic systems.
• Strong, hands-on experience building production agent systems with modern orchestration.
• Experience designing and optimizing end-to-end RAG systems.
• Strong understanding of memory and context management.
• Deep, practical understanding of LLM behavior, including strengths, limitations, hallucination risks, reasoning constraints, and latency/cost trade-offs, and evaluation methods.
• Experience evaluating and debugging agent behavior at task and trajectory levels.
• Familiarity with vector databases and retrieval infrastructure (e.g., Pinecone, Weaviate, or Milvus) is preferred.
• Exposure to model adaptation and fine-tuning techniques (e.g., LoRA or QLoRA) is preferred.
• Understanding of traditional NLP concepts (tokenization, semantic similarity, entity extraction, summarization, transformer fundamentals) is preferred.
• Experience operating in highly regulated, high-stakes, or operationally complex environments; healthcare exposure is a plus, not a requirement.
• Demonstrated habit of staying current with AI research, benchmarks, and emerging engineering patterns.