JevMade field notes / Architecture and evaluation walkthrough
RazorRisk — Agentic AI Payment Fraud & Risk Investigation Platform
A long-form walkthrough of transaction scoring, graph-based fraud-community detection, learned score fusion, live investigation, evaluation, and the limits of what the supplied evidence establishes.
Original by Dikshu015EvaluationRepository READMESource reviewed
Before you dive in
What you’ll find in the original
Keep transaction classification, graph signals, score fusion, and the Jev verification step explicit rather than presenting one opaque risk score.
Document the training and live-transaction contracts separately so an offline evaluation is not mistaken for a production guarantee.
Pair quick-start and demo instructions with a limitations section that distinguishes demonstrated behavior from unproved operational claims.
Worth knowing
The evaluation and system claims are reported by the repository author and were not independently reproduced.