Agentic AI Systems
Scamminator - Fraud Detection Assistant
A multi-agent, LLM-powered assistant that evaluates e-commerce products and sellers for trustworthiness, providing verdicts, scores, and actionable insights to help users make safer purchases.
Explore the Project
Problem
Online shoppers struggle to identify fraudulent products, fake reviews, and untrustworthy sellers because critical trust signals are scattered across product listings, reviews, and seller profiles.
Solution
Developed Scamminator, a multi-agent LLM-powered fraud detection assistant that investigates product descriptions, reviews, and seller information to generate explainable trust scores, scam likelihood assessments, and user recommendations.
Architecture
Next.js and Streamlit frontend integrated with Python-based investigator agents, Selenium scraping pipeline, Gemini-powered reasoning layer, and a final judge agent that aggregates evidence into a unified fraud assessment.