Ai-powered Compliance Monitoring System
Link to open source: https://github.com/abhay-2108/ComplainceAI
ComplianceAI is an AI-powered Anti-Money Laundering (AML) compliance platform that automates the entire transaction monitoring pipeline — from ingestion and risk scoring to AI-generated compliance reports and human review.
Why we built it: Traditional AML systems rely on rigid if-else rules that criminals easily exploit by splitting large transfers into smaller ones below detection thresholds. Analysts then spend 30–45 minutes manually reviewing each alert, writing explanations, and deciding whether to escalate — while thousands of alerts pile up daily. We built ComplianceAI to eliminate this bottleneck entirely.
How it works: A Random Forest classifier (99% accuracy, trained on real AML transaction data) flags suspicious transactions based on 15 engineered features — including cross-currency patterns, transaction velocity, and timing anomalies. Every flagged violation is then explained by Google Gemini 2.5 Flash using a RAG pipeline grounded in real AML policy documents, generating a professional, auditor-ready compliance report in under 5 seconds. Compliance officers review each flag via a Human-in-the-Loop panel, where they can Resolve or Escalate with notes — creating a full, timestamped audit trail.
What makes it different: The entire system runs on a distributed Celery task queue backed by MongoDB Atlas, processing batches of 5–50 transactions in parallel without blocking the UI. Five specialized AI agents handle monitoring, detection, explanation, policy retrieval, and reporting — end to end, automatically.
The platform ships with a 9-page React dashboard covering real-time KPIs, live agent activity, violation explorer, ML prediction analytics, policy management, downloadable compliance reports, and audit logs.
This build was uploaded as a hackathon project
