
AI-powered dual-stack fraud detection system (Python ML & TypeScript API Gateway) using Logistic Regression, TensorFlow, and LLaMA models with Convex.dev for scalable data persistence.
This project, developed during HackMIT 2024, is a robust backend for managing user transactions and detecting potential fraud. It uses a dual-stack architecture: a Python FastAPI service for machine learning-driven fraud prediction (Logistic Regression, TensorFlow, LLaMA-based analysis) and a TypeScript Express.js server as an API gateway for routing and data persistence with Convex.dev. It features user management, comprehensive transaction handling, asynchronous fraud processing, and Dockerized deployment. The system acts as an orchestrator, delegating fraud prediction to the Python ML service and updating transaction status in Convex.