Vigil360 — KYC & Continuous Risk Screening

5th place out of 23 teams at InterIIT TechMeet 14.0, IIT Patna

The problem statement asked teams to build an end-to-end financial product on top of Pathway capable of handling high-velocity streaming data. We built Vigil360 - a KYC screener with continuous risk evaluation that combines live sanctions/adverse-media data with transaction-pattern matching to flag suspicious entities in real time with enriched context of the person from the KYC and online information seamlessly integrated into a compliance dashboard.

What it does

• Ingests KYC form data and runs it through a multi-signal risk pipeline (sanctions lists, adverse media, transaction behavior)
• Assigns a Risk Propensity Score (RPS) using an ensemble of rule-based and ML models (CatBoost, logistic regression)
• Streams updates via Kafka + Pathway so risk scores update continuously — not just at onboarding
• Surfaces flags and explanations through a Next.js dashboard for compliance teams

Architecture

Stack

• Pipeline — Pathway, Apache Kafka, PostgreSQL, Debezium, Docker, Pathway MCP Server
• Backend — FastAPI, SQLAlchemy
• Frontend — Next.js 15, React 19, TypeScript, Tailwind CSS
• ML — CatBoost, logistic regression ensemble, RAG (ChromaDB + Mistral)
• Data sources — OpenSanctions API, adverse media scraping with news source authenticity judging, OFAC lists

Getting Started

git clone https://github.com/Quantum-Codes/Pathway-InterIIT-14.git
cd Pathway-InterIIT-14

Start the three components in order:

1. Backend

cd backend
# See backend/README.md

2. Pipeline

cd pipeline
# See pipeline/README.md — sets up Kafka, PostgreSQL, Pathway streams

3. Frontend

cd frontend
# See frontend/README.md