Drop-in descriptions for resume bullets, healthcare-targeted variants, and LinkedIn-length copy. All variants describe the same project (Scribe IQ product case study); pick the one that matches the audience.
Scribe IQ — healthcare AI platform prototype. Built an end-to-end demonstration of data-product style synthetic clinical corpus construction, governed AI serving, and clinical documentation workflows. The system separates offline corpus construction from runtime serving: a nine-step data_prep/ pipeline builds a validated synthetic corpus artifact from Synthea and public clinical note sources; FastAPI/Postgres/pgvector serve patient charts, encounter views, grounded RAG chat with citations, structured note generation, and an ai_interactions audit dashboard. Pluggable LLM/embedding providers (Groq, OpenAI, Azure OpenAI, Amazon Bedrock); synthetic data only with explicit PHI, SSO, tenancy, and BAA caveats.
Scribe IQ — grounded clinical documentation demonstration. Designed and built an end-to-end clinical AI documentation system (Next.js, FastAPI, Postgres + pgvector) to demonstrate the architectural restraint and governance posture required for institutional healthcare AI. Treats hallucination as a safety failure rather than a usability bug: every chat answer is retrieval-grounded with a [note:uuid] citation contract enforced in the system prompt, audited AI paths write append-only rows to a first-class ai_interactions table with admin-visible success, degraded, failed, and blocked status aggregation, and provider boundaries (Groq demo, Azure OpenAI / Amazon Bedrock for institutional deployments) are stated explicitly with clear caveats that enterprise providers do not by themselves create PHI compliance. Bridges institutional data-system experience from higher-education (longitudinal records, governance as schema, multi-stakeholder views) into clinical-shaped systems. Synthetic corpus only; documentation names what would change for a production PHI deployment (SSO, multi-tenant isolation, BAA, formal de-identification, observability).
Built Scribe IQ, an end-to-end grounded clinical documentation demo — Next.js + FastAPI + Postgres/pgvector with citation-contract RAG, first-class ai_interactions audit, and pluggable LLM/embedding providers (Groq / OpenAI / Azure OpenAI / Amazon Bedrock). Synthetic data only; the goal is to show how I think about governance, provider boundaries, and what to leave out — not to ship a clinical product. Case study: PORTFOLIO_CASE_STUDY.md.