Vincere.dev Vincere
Healthcare Production System

Meditap

Healthcare Data Warehouse Platform

Meditap
60+
Hospitals Served
~1 min
Data Latency
100+
Pipelines Deployed
0
Database Outages

Executive Summary

We built an event-driven Google Cloud data warehouse. CDC, Pub/Sub, and BigQuery separated analytics from transactional systems. Database downtime ended, and data latency reached about one minute across 60+ hospitals.

The Problem

Meditap needed a scalable platform for 60+ private hospitals in Indonesia. Analytics ran on the primary database, causing 3–5 outages daily. A 3–5 person data team needed near real-time reporting, consistent distributed data, safe historical migration, governance, and continuous ingestion.

60+
Hospitals
3–5
Daily Outages
3–5
Data Team Size
Services Delivered
AI Integration Dedicated Team

Healthcare Data Warehouse Platform

Architecture Overview

Data Layer
PostgreSQL Microsoft SQL Server Debezium Google Cloud Pub/Sub
Backend & Orchestration
FastAPI Apache Airflow
Frontend
React
Infrastructure
GCP GKE BigQuery

Key Technical Decisions

System Design

The platform is layered and event-driven. PostgreSQL and SQL Server are sources. Debezium captures changes through log-based CDC and streams them through Pub/Sub. BigQuery stores raw data, while Airflow orchestrates ETL. A custom UI compiles YAML configurations into DAGs. Dataset-level IAM controls access. Each layer can scale and fail independently.

Key Decisions

CDC replaced batch ingestion to make data available in about one minute. A template-driven Airflow platform removes manual DAG coding. Non-engineers can operate pipelines. BigQuery provides serverless analytics with low infrastructure overhead. The tradeoff is less room for custom pipeline logic.

Implementation Highlights

Debezium provides reliable change tracking. Pub/Sub decouples ingestion. The warehouse separates raw, processed, and reporting-ready data. Platform validation blocks bad configurations before DAG generation. Near real-time ingestion pairs with batch transformations for aggregation.

Results & Validation

Eliminated primary database downtime, previously 3–5 outages daily.

Reduced report generation from minutes to seconds.

Reached about one-minute data latency.

Enabled a 3–5 person team to run the full pipeline ecosystem.

Deployed and operated 100+ pipelines through the platform.

Key Insights

About one-minute data freshness without overloading source systems.

Pipelines created through the platform without code.

100+ pipelines managed through reusable templates.

Transactional and analytical workloads decoupled to remove contention.

Dataset-level IAM provides role-based access.

Who This Applies To

Applicable to high-volume transactional systems that need real-time analytics without risking operations. It fits healthcare, fintech, and multi-tenant enterprise environments.

Healthcare Data Warehousing Real-Time Analytics CDC & Streaming Platform Engineering

Technologies Used

Backend

FastAPI React

Frontend

GCP

Infrastructure

GKE PostgreSQL Google Pub/Sub

Data & Integrations

Microsoft SQL Server Apache Airflow BigQuery

Patterns & Techniques

Debezium CDC YAML DAGs IAM Bitbucket

Tools

Jira Keycloak

Building something similar?

We specialize in ai integration and dedicated team for healthcare companies. If you're facing challenges like the ones we solved for Meditap, let's talk.

30 minutes · No commitment · We respond within 24h