Final Year Project

Project Title

AI Operations Agent for Payment Systems

Description

This project developed an AI Operations Agent for TPS to automate incident resolution in high-stakes payment systems. By orchestrating LangGraph with a human-in-the-loop design, the agent automates the entire lifecycle from log parsing to Pull Request creation. The technical core is a Hybrid RAG system (BM25 + FAISS) that maps complex transaction errors to source code with high precision using GPT-4o. Built on a high-concurrency FastAPI backend with WebSockets, it provides real-time visibility into the agent's reasoning. The primary benefit is a drastic reduction in Mean Time to Repair (MTTR). By automating routine debugging, the system increases operational efficiency, minimizes financial risk from downtime, and frees engineers from manual log analysis.