A production-grade, AI-powered fraud operations platform built for modern fintechs, banks, and payment processors.
This platform bridges the gap between raw Machine Learning outputs and human Compliance Operations, providing a complete 360° lifecycle for fraud intelligence: from real-time ML inference, heuristic business rules, and Explainable AI (XAI) to Case Management, Data Drift Monitoring, and Executive Analytics.
- Predictive Engine: High-performance Random Forest model for transaction scoring.
- Explainable AI (XAI): Real-time SHAP-like feature contributions (waterfall charts) for deep transparency.
- Model Registry: Version control for ML models with one-click production promotion/rollback.
- Data Drift Monitoring: Continuous Population Stability Index (PSI) tracking to detect feature drift over time.
- Business Rules Engine: Dynamic heuristic rules (e.g., Velocity, Impossible Travel) that override or augment ML scores.
- Case Management System: Full investigative queue with timelines, internal notes, and assignment (Open, Investigating, Escalated, Resolved).
- Role-Based Access Control (RBAC): Secure access tiers (Admin, Analyst, Executive, Data Scientist).
- Immutable Audit Logs: Strict tracking of every action (status changes, model deployments, rule creation) for regulatory compliance.
- Customer Risk 360: Lazy-aggregated profiles showing lifetime risk scores, fraud flags, and mock device/location intelligence.
- Executive Dashboard: High-level KPIs measuring revenue saved (blocked fraud) vs potential loss (approved fraud), and SLA tracking.
- False Positive Analytics: Analyst leaderboard and operational waste projections to optimize alert accuracy.
- Live Simulator: Real-time bursty transaction generator simulating WebSocket gateway feeds for velocity monitoring.
- Compliance Export Center: CSV, Excel, and JSON data extraction for regulatory audits.
- Python 3.10+
- 8GB RAM minimum (Optimized for standard dev environments)
- Clone the repository and install dependencies:
pip install -r requirements.txt- Generate Demo Data & Migrations (Enterprise Seeder):
python scripts/seed_enterprise_data.py(This idempotently builds the DB schema, seeds 25 historic cases, 50 audit logs, 8 active business rules, and creates 4 demo users).
- Run the Platform:
python run.py(This launches both the FastAPI backend on port 8000 and the Streamlit UI on port 8501).
Access the portal at http://localhost:8501 using one of the seeded accounts:
- Admin/Exec:
admin/admin123 - Data Scientist:
ds_user/ds_123 - Senior Analyst:
analyst_sr/analyst_123 - Junior Analyst:
analyst_jr/analyst_123
Built as a showcase for production-level ML Engineering, Full-Stack Architecture, and DevOps.