AI Risk Manager: Credit Card Fraud Detection

Production-grade fraud detection system for Indian BFSI targeting credit card and transaction fraud with measured precision and recall.

Overview

Stop merchants losing money to fraud with an AI-powered fraud detector built on real Indian transaction patterns.

Key Features

• ✅ 54.21% Fraud Recall - Catches majority of fraudulent transactions
• ✅ ₹2.61 Crores Net Business Value - Quantifies fraud prevented vs false positive costs
• ✅ Indian BFSI Context - Realistic merchants, amounts in ₹, IST timezone
• ✅ Production Ready - Flask API with 4 endpoints, Docker deployment
• ✅ Transparent Evaluation - Full clarity on precision/recall trade-offs

Model Performance

Metric	Value
Precision	10.82%
Recall	54.21%
F1-Score	0.1804
ROC-AUC	0.6920

Business Impact

• Fraud Caught: 2,732 transactions
• Fraud Value Prevented: ₹2,73,20,000
• False Positive Cost: ₹11,25,750
• Net Business Value: ₹2,61,94,250

Quick Start

pip install -r deployment/requirements.txt
python src/api.py

API Endpoints

• GET /health - Health check
• GET /info - Model metadata
• POST /predict - Single/batch prediction
• POST /batch-predict - Batch with statistics

Model Details

• Algorithm: XGBoost
• Features: 42 (PCA + engineered)
• Training: 70-30 stratified split

Presentations

AI Risk Manager Pitch Deck (View Online)

• 11-slide overview for recruiters and investors
• View online or download

Contents:

• Credit card fraud problem (₹2,000+ Cr annually)
• AI Risk Manager solution (42 features, XGBoost)
• Business results (₹2.61 Cr profit)
• Technical architecture
• 2 AM debugging story
• Roadmap for improvements

Pitch Recording

https://drive.google.com/uc?id=1YpEv-Qw6xz8B5edeNuD73Tk2bJSb9JYD&export=download