VULCAN SHIED | RAZORPAY | AI BUILDATHON

Vulcan Shield AI-Powered Payment Risk, Revenue & Optimization Intelligence Platform Github : https://github.com/V3DxNT/VulcanShield What if payment systems could do more than simply approve or reject transactions? This project presents an AI-powered unified payment intelligence and decisioning platform designed to analyze transactions in real time, detect fraud and anomalies, protect revenue, and optimize payment success while minimizing unnecessary friction for legitimate customers. The system combines real-time transaction analytics, supervised fraud detection, anomaly detection, behavioral intelligence, RAG-based contextual analysis, fraud-network intelligence, and AI-assisted investigation to make smarter payment decisions. 🔍 Key Capabilities • ⚡ Real-Time Transaction Intelligence — continuously analyzes incoming payment events. • XGBoost Fraud Detection — identifies transaction patterns associated with known fraud. • Isolation Forest Anomaly Detection — detects unusual transaction behavior that may not match known fraud patterns. • 🧠 Behavioral Intelligence / User History RAG — compares transactions against historical customer behavior. • 🌐 Global Fraud Intelligence — retrieves relevant patterns and historical fraud knowledge using RAG/vector search. • Fraud Graph Analysis — identifies suspicious relationships between users, devices, IPs, and other entities. • Risk-Based Decision Engine — dynamically chooses between ALLOW, CHALLENGE, and BLOCK. • Challenge-Based Verification — uncertain transactions can be verified instead of being immediately rejected. • 💬 AI-Assisted Investigation — Qwen2.5 7B via Ollama provides contextual investigation and explanations using evidence gathered from the system. • Event-Driven Architecture — Go, Kafka, Redis, PostgreSQL, FastAPI, and real-time socket communication work together to process transactions efficiently. 🎯 What Are We Trying to Achieve? The fundamental challenge in payment systems is balancing fraud prevention with payment conversion. Blocking too aggressively can prevent legitimate customers from completing payments, while being too permissive can increase fraud and revenue leakage. The goal of this platform is therefore not to block more transactions — but to make better payment decisions. By combining multiple layers of intelligence, the system aims to: ✅ Detect fraudulent transactions ✅ Identify previously unseen anomalies ✅ Reduce false positives ✅ Minimize unnecessary customer friction ✅ Protect legitimate payment revenue ✅ Improve payment success and safe conversion ✅ Detect coordinated fraud patterns ✅ Provide contextual and explainable risk decisions ✅ Make real-time, evidence-based payment decisions 🏗️ Technology Stack Backend: Go Event Streaming: Apache Kafka Database: PostgreSQL Real-Time State: Redis ML Service: FastAPI ML Models: XGBoost + Isolation Forest AI: Ollama + Qwen2.5 7B Intelligence: RAG + Vector Search + Fraud Graph Frontend: Next.js Communication: WebSockets / Socket Broadcasting Architecture: Event-driven microservices 💡 Core Philosophy Maximum safe conversion — not maximum blocking. A payment that looks suspicious shouldn’t always be rejected. Sometimes the right decision is to challenge it, gather additional evidence, and allow it if the customer can successfully verify the transaction. This project explores how AI, machine learning, real-time event processing, behavioral intelligence, and network-level analysis can work together to build a more intelligent payment decisioning system. Built as an independent experimental project for exploring AI-powered payment intelligence and risk management. ⸻ 🔖 Hashtags #AI #ArtificialIntelligence #MachineLearning #FraudDetection #FraudPrevention #PaymentSecurity #PaymentTechnology #FinTech #FintechInnovation #PaymentGateway #RiskManagement #RiskEngine #AnomalyDetection #XGBoost #IsolationForest #RAG #GenerativeAI #LLM #Qwen #Ollama #VectorDatabase #FraudAnalytics #CyberSecurity #DataScience #BackendDevelopment #GoLang #Golang #ApacheKafka #Kafka #Redis #PostgreSQL #FastAPI #NextJS #WebSockets #Microservices #EventDrivenArchitecture #SoftwareEngineering #AIEngineering #BuildInPublic #Hackathon #Razorpay #RazorpayBuildathon #AIInnovation #TechProject