Mortality Signals — AI-Powered Global Mortality Analytics

Transform global mortality data into actionable intelligence through AI-powered anomaly detection, scenario modeling, and interactive Tableau Cloud visualizations for public health decision-making.

Comment

Inspiration

Public health analysts and policymakers face a critical challenge: sifting through vast amounts of global mortality data to identify where interventions are most needed. They need answers to pressing questions: Where are unusual spikes in deaths occurring? What's causing them? And most importantly, what interventions would save the most lives? Traditional analysis is time-consuming and often reactive rather than proactive.

What it does

Mortality Signals transforms 61 countries' mortality data across 30 causes into actionable intelligence through:

  • AI-Powered Signal Detection: Z-score anomaly detection automatically surfaces 4,500+ mortality anomalies with severity levels (Critical/Warning/Info)
  • Interactive What-If Scenarios: Scenario Builder allows users to model intervention strategies, showing projected lives saved with specific reduction targets
  • Tableau Cloud Integration: JWT-authenticated embedded dashboards provide deep analytical views with Top 10 Causes, yearly trends, and interactive filters
  • Global Observatory Dashboard: Real-time KPIs, 30-year trends, and cause category breakdowns in a professional dark/light themed UI
  • Peer Country Comparison: Compare up to 6 countries with indexed/absolute views to identify best practices

How we built it

Data Pipeline (Python):

  • ETL pipeline processing Kaggle mortality data (1990-2020)
  • Z-score statistical analysis for anomaly detection
  • Parquet file optimization for fast querying

Backend (FastAPI):

  • RESTful API with endpoints for data, insights, scenarios, and Tableau integration
  • JWT token generation for secure Tableau Cloud embedding
  • Scenario simulation engine with lives-saved calculations
  • CORS-enabled for production deployment

Frontend (React + TypeScript):

  • Modern SPA with Vite build system
  • Tableau JavaScript API v3 integration
  • Responsive design with Tailwind CSS
  • Dark/light theme support

Tableau Cloud:

  • Connected App with Direct Trust authentication
  • Published dashboards with global mortality visualizations
  • Embedded views with row-level security

Deployment:

  • Google Cloud Run for API and web services
  • Docker containerization
  • Production URLs for live demo

Challenges we faced

  • Tableau JWT Authentication: Implementing secure server-side JWT token generation while keeping secrets safe from client exposure
  • Data Volume: Processing and optimizing 30 years of global mortality data for real-time queries
  • Anomaly Detection Tuning: Calibrating Z-score thresholds to surface meaningful signals without overwhelming users with false positives
  • Scenario Modeling Logic: Designing an intuitive intervention builder that accurately projects lives-saved outcomes
  • Cross-origin Integration: Configuring CORS and Tableau's domain allowlist for seamless embedding

Accomplishments that we're proud of

  • Built a production-ready platform with 4,500+ automatically detected mortality anomalies
  • Achieved seamless Tableau Cloud integration with JWT authentication
  • Created an intuitive Scenario Builder that makes complex what-if analysis accessible
  • Developed a polished, professional UI that works in both dark and light modes
  • Deployed live demo accessible at production URLs
  • Delivered a complete solution from data pipeline to frontend in hackathon timeframe

What we learned

  • Deep understanding of Tableau Embedded Analytics and Connected Apps architecture
  • Best practices for JWT-based authentication in embedded analytics
  • Statistical methods for detecting meaningful anomalies in time-series health data
  • Importance of user experience in making complex analytical tools accessible
  • Google Cloud Run deployment strategies for full-stack applications

What's next

  • Real-time data streaming with Kafka for immediate anomaly alerts
  • Advanced forecasting using Prophet/ARIMA models
  • User authentication with OIDC for personalized dashboards
  • Alert subscriptions and email notifications for critical signals
  • Tableau Pulse integration for mobile insights
  • Kubernetes deployment for enhanced scalability
  • Integration with WHO and CDC data sources
Share this project:

Updates