Derrick-Ryan-Giggs/kenya-health-pipeline

Open-source lakehouse mapping Kenya health facility distribution across 47 counties — Airflow, MinIO, Iceberg, Trino, dbt Core, Superset # Kenya Health Facility Mapping Pipeline > **An open-source data lakehouse exposing healthcare inequality across Kenya's 47 counties — built end-to-end with Apache Airflow, MinIO, Apache Iceberg, Trino, dbt Core, and Apache Superset, all running in Docker on a local machine.** --- ## Table of Contents - Problem Description - Solution Overview - Architecture - Tech Stack - Data Sources - Project Structure - Infrastructure as Code (OpenTofu) - Data Ingestion & Orchestration - Data Lakehouse Design - Transformations (dbt Core) - Dashboard - Challenges & How We Solved Them - Reproducibility — How to Run - Known Limitations & Future Work - References - Author --- ## Problem Description Kenya has 47 counties and over 20,000 registered health facilities — yet access to care is profoundly unequal. A child born in Westlands, Nairobi, has access to 7.23 facilities per 10,000 people. A child born in Embakasi North has 0.62. That inequality is real, large, and invisible — because the data to prove it has never been systematically assembled in one place. The Ministry of Health (MOH) publishes facility lists through the Kenya Master Health Facility Registry (KMHFR). The Kenya National Bureau of Statistics (KNBS) publishes county population data from the 2019 Census. The Humanitarian Data Exchange (HDX) publishes county boundary GeoJSON. But no one has joined these three datasets together into a reproducible, automated pipeline that answers the real questions: - **Which counties are most underserved relative to their population?** - **Which counties lack maternity, ART (HIV treatment), TB, or emergency facilities?** - **How has the facility landscape changed over time?** - **What would it take to bring every county to a baseline of 3 facilities per 10,000 people?** This pipeline answers all of those questions — automatically, monthly, and at county and sub-county resolution — using 100% open-source tools running in Docker on a local machine. --- ## Solution Overview | …