Jayesh-R/Ghana-Healthcare-Intelligence-Platform

Databricks Accenture Hackathon # 🏥 Ghana Healthcare Intelligence Platform ### Databricks × Accenture Hackathon 2026 — Virtue Foundation Track > **Bridging Medical Deserts with Agentic AI** — An intelligent document parsing and resource planning system that transforms 987 messy Ghana hospital records into actionable healthcare intelligence for NGO planners. --- ## 🎯 What It Does The Ghana Healthcare Intelligence Platform is a full-stack agentic AI system that: - **Finds medical deserts** — regions in Ghana with zero emergency, ICU, or surgical care - **Parses unstructured hospital data** — converts messy Facebook posts and web listings into structured medical schema - **Detects anomalies** — flags hospitals with suspicious or inconsistent capability claims - **Plans resource deployment** — AI-powered recommendations on where to send doctors - **Answers natural language queries** — NGO planners can ask questions in plain English and get cited, traceable answers --- ## 🚀 Live Demo | Credential | Value | |------------|-------| | Demo URL | `localhost` (local) | | Test Email | `demo@virtuefoundation.org` | | Test Password | `demo1234` | > Register a new account. --- ## 📸 Screenshots ### Dashboard — Real-time stats from 987 hospitals - 847 total facilities across Ghana - 23 medical deserts identified - 3 critical regions with zero emergency care - 58% average data completeness ### AI Agent — RAG-powered natural language queries - Ask: *"Which regions have no emergency care?"* - Get: Real hospital names, citations, agent reasoning steps - Powered by: FAISS semantic search + Groq/Llama-3.1 ### Medical Deserts Map — Interactive Ghana coverage map - Color-coded by gap severity (green/yellow/red) - Click any region for detailed breakdown - Identifies Upper West, Upper East, Northern as critical ### Anomaly Detection — 50 flagged suspicious records - Hospitals claiming surgery with no equipment listed - Facilities with ICU claims but zero bed capacity - Incomplete records needing v …