Overview
Built an event-driven healthcare risk monitoring pipeline that simulates insurance claim events and cybersecurity access logs, detects suspicious behavior in real time, and correlates provider-level alerts in a unified dashboard.
Problem
Healthcare risk operations often separate fraud events from cybersecurity activity, which makes it harder to detect emerging provider-level risk across data sources and alert streams.
What I Built
- Simulates fraud-oriented claim events and cybersecurity access-log events.
- Uses producers and consumers in a Kafka or Confluent-style architecture.
- Applies rule-based fraud checks and cyber anomaly detection logic.
- Correlates signals into a unified provider risk view.
- Stores alerts in SQLite and surfaces them in a Streamlit dashboard.
- Supports a local demo mode without Kafka for easier review and testing.
System Architecture
- Event producers for claims and cyber activity.
- Consumer services for fraud rules and cyber anomaly monitoring.
- Unified provider risk correlation logic.
- Alert persistence layer for auditability and review.
- Streamlit presentation layer for operational visibility.
Business Value
This earlier healthcare risk project demonstrates streaming analytics, data engineering, and event-driven alerting in a domain where fraud and cyber signals often intersect.
Legacy Positioning
This project is intentionally preserved as a companion legacy system to the newer AI-powered healthcare fraud decision-support platform. Together, they show both:
- streaming and event-driven risk detection
- machine learning and business decision support
Future Hybrid Roadmap
Future versions may combine the streaming event pipeline with the machine learning fraud scoring dashboard to create a unified real-time healthcare risk platform.
Limitations
The system is intentionally preserved as a legacy companion project, so the public case file emphasizes architecture and signal correlation rather than a newer machine-learning workflow.