Banking & Finance
Real-Time AI Fraud Detection Platform for Tier-1 Bank
Architected and deployed a low-latency, real-time machine learning system to prevent payment fraud across 15 markets.
Completed: 11/15/2024
🎯 Project Scope
Full-cycle platform build including model training pipeline, API development, and cloud deployment on a secured VPC.
⛰️ The Challenge
Existing rule-based system could not adapt to new fraud vectors, resulting in significant annual losses and high manual review overhead.
🛠️ The Hanva Solution
Implemented a Graph Neural Network (GNN) model and deployed it via a serverless, event-driven architecture (Kafka, Lambda) for immediate transaction scoring.
Key Technologies & Architecture
- • TensorFlow
- • Kubernetes
- • AWS EKS
- • Kafka
- • Python