Principal AI Architect

We are working with a leading healthcare revenue cycle technology company that is looking for a Principal AI Architect to join its team. This is a newly created, hands‑on individual contributor role that owns the AI technical function at the company: the technical vision, the architecture, and the quality and cost of everything the company builds with AI. The successful candidate will be the company's deepest AI technical authority, building its AI, automation, and intelligence capabilities in‑house on a centralized, modern lakehouse data platform.

The role is expected to evolve into Head of AI Engineering as the team grows.

• Set the technical vision and roadmap with the executive leading AI and automation, giving leadership a clear view of feasibility, value, cost, and risk.
• Decide what should be deterministic (rules, parsers, schema validation, SQL, classical ML) and what should be dynamic (LLM reasoning, agentic tool use), using LLMs where inputs are ambiguous and plain code everywhere else.
• Design single- and multi‑agent systems: routing, planner/executor patterns, tool calling and MCP, state and memory, durable execution, retries, idempotency, human approval gates, and tool permission scoping.
• Build and tune RAG end to end: parsing, chunking strategy, embedding selection and fine‑tuning, hybrid search, re‑ranking, query rewriting, vector search, and grounded citations.
• Build knowledge graphs and ontology‑driven retrieval, including entity resolution and GraphRAG‑style hybrids, in partnership with the data organization.
• Handle document intelligence for contracts, payment files, and correspondence: OCR, layout‑aware parsing, table and field extraction, and multimodal models.
• Design the data layout AI depends on: medallion architecture, Delta table design, Unity Catalog governance and lineage, and feature and serving tables.
• Select models by task, accuracy, latency, cost, and data‑handling constraints. Fine‑tune where it pays off (SFT, LoRA/QLoRA, preference tuning) and distill into smaller, cheaper models.
• Deploy and serve on managed model serving and self‑hosted inference such as vLLM, using quantization, batching, autoscaling, and routing. Run the lifecycle in MLflow with shadow, canary, and A/B rollouts.
• Build evaluation and monitoring tied to business outcomes: golden datasets, retrieval metrics, groundedness, LLM‑as‑judge calibrated to human labels, release‑gating regression suites, and production drift monitoring.
• Own token and compute spend: routing and cascading, prompt and semantic caching, context compression, batch inference, and cost attribution per agent and workflow.
• Defend against prompt injection, tool‑based data exfiltration, and over‑permissioned agents with least privilege, guardrails, PII/PHI redaction, and full audit logging.
• Lead the technical build from design through production, writing the specs, patterns, and standards everyone building AI at the company works to, and mentor the AI engineering team that forms around this role.

• 10+ years in software, data, or ML engineering, with 7+ years in machine learning and increasing scope.
• Strong current Python and SQL. Builds, tests, and debugs AI systems personally.
• Production experience across LLM agents, RAG, knowledge graphs, closed and open‑weight models, fine‑tuning, and deployment.
• A record of proving ROI and controlling AI cost, and of choosing deterministic versus LLM‑driven design well.
• Experience securing AI systems in regulated or sensitive‑data environments.
• A record of setting technical direction others executed, including across employees, contractors, and vendors.

• Healthcare, healthcare IT, or revenue cycle management (RCM) background, including HIPAA and…