Fraud Analytics and Bayesian Risk Decisioning Platform

A full-stack fraud analytics platform combining machine learning, Bayesian risk decisioning, and modern dashboards.

Tech Stack

• Frontend: Next.js 14, TypeScript, TailwindCSS, Recharts
• Backend: Node.js, Express, TypeScript
• Data: BigQuery, Google Cloud Storage
• ML: Vertex AI (training & serving), BigQuery ML
• Risk Engine: Custom Bayesian module

Architecture

graph TD
    User([User]) -->|HTTPS| Frontend[Next.js Frontend]

    subgraph "Frontend Layer"
        Frontend
        Dash[Dashboards] -->|Visualizes| Metrics[KPIs & Metrics]
        RiskLab[Risk Lab] -->|Interactive| Sim[Simulation & Thresholds]
    end

    Frontend -->|API Requests| API[Backend API Node.js]

    subgraph "Backend Layer"
        API
        Bayesian[Bayesian Risk Engine]
        API <-->|Posterior & Cost| Bayesian
    end

    subgraph "Data & ML Infrastructure"
        BQ[(BigQuery)]
        Vertex[Vertex AI]
        GCS[(Google Cloud Storage)]
    end

    API -->|SQL Queries| BQ
    API -->|Prediction / GenAI| Vertex
    Vertex -->|Training Data| BQ
    Vertex -->|Artifacts| GCS

Loading

Project Structure

├── frontend/          # Next.js application
├── backend/           # Node.js API server
├── ml/                # ML training and deployment
│   ├── data_pipeline/ # Data ingestion and feature engineering
│   ├── notebooks/     # Exploratory analysis
│   ├── training/      # Model training scripts
│   ├── evaluation/    # Model evaluation
│   └── deployment/    # Vertex AI deployment
├── infrastructure/    # GCP setup scripts
└── docs/             # Documentation

Quick Start

Backend

cd backend
npm install
cp .env.example .env
# Edit .env with your GCP credentials
npm run dev

Frontend

cd frontend
npm install
cp .env.example .env.local
# Edit .env.local with your API URL
npm run dev

Features

• ML Fraud Detection: XGBoost model trained on 6.3M transactions
• Bayesian Risk Engine: Sophisticated probabilistic decisioning
• Three Governance Modes: Strict, Balanced, Precision
• Real-time Dashboards: KPIs, analyst queue, supervisor metrics
• Risk Lab: Interactive threshold exploration and sequential testing
• AI Insights: Vertex AI-powered narrative insights

Development Status

See task.md for detailed progress tracking.

License

ISC