Skip to main content

ZOL Intelligent Search

Transforming hospital search from keyword matching to semantic understanding. A Retrieval Augmented Generation system that helps 100,000+ monthly visitors find the care they need.

Explore the ArchitectureWhat is RAG?Built with the S4U Methodology
Monthly Visitors
Search Queries / Month
Medical Brochures
Conditions Covered

Semantic Search

Retrieval Augmented Generation replaces brittle keyword matching with deep semantic understanding. Queries in natural Dutch are transformed into vector embeddings and matched against verified hospital content.

Knowledge Graph

A PostgreSQL-backed taxonomy captures the relationships between doctors, departments, conditions, treatments, and campuses -- enabling structured entity lookups that pure vector search cannot achieve.

Medical Safety

A multi-layer safety architecture ensures the system never provides medical advice. Intent classification, post-generation validation, quality gates, and mandatory disclaimers work in concert.

How this was built

S4U Development Methodology​

AI-Assisted Software Engineering at Production Quality

This system was not vibe-coded. It was built under a documented engineering methodology that treats the AI assistant as a senior engineer working under architectural oversight -- with the quality bar enforced by infrastructure rather than by discipline.

Design before codeEvery non-trivial change goes brainstorm to design to execution, with one committed design artifact.
Gates are mechanismsEvery quality gate names an enforcing hook, CI check, or script. "The agent will remember" is not a mechanism.
Verification before completionNothing ships on "it should work now". Fresh verification output, or it is not done.
Decisions are recordedTechnology, architecture, and safety choices land as ADRs with a register checked in CI.
Read the methodology

Built for Healthcare

The ZOL Intelligent Search system combines multiple AI disciplines into a cohesive, safety-first architecture designed specifically for the healthcare domain.

  • Retrieval Augmented Generation grounds every response in verified hospital content
  • Knowledge graph captures relationships between doctors, departments, conditions, and treatments in the PostgreSQL taxonomy
  • Multi-layer safety ensures the system never provides medical advice
  • Hybrid evaluation validates response quality in real time
View System Architecture
Frontend
React + TypeScript chat interface
API Layer
FastAPI with WebSocket streaming
RAG Pipeline
Intent, retrieval, generation, evaluation
Data Layer
PostgreSQL + pgvector + Redis + MinIO