About Ghost Factory
Ghost Factory was inspired by the challenge of detecting coordinated financial fraud without unfairly flagging legitimate borrowers. We wanted to build an AI system that investigates suspicious connections rather than simply treating them as proof of fraud.
We built an autonomous auto-loan fraud investigator using Python, Nicrron, and PRISM by Block Convey. The system analyzes relationships between applications, retrieves employment and document verification records, and generates evidence-backed recommendations. We also developed an interactive dashboard with a fraud network visualization and conversational AI assistant.
One of our biggest challenges was distinguishing genuine fraud indicators from innocent connections, such as unrelated applicants using the same dealership kiosk. We also learned how to integrate AI tool calling, enforce financial safety guardrails, and validate AI-generated findings.
Using a synthetic dataset of 12 applications, Ghost Factory successfully retained all five fraudulent cases for review while clearing one false-positive alert. Our key takeaway was that trustworthy financial AI should not just detect risks, but also question its own findings and know when to leave decisions to humans.## Inspiration
Log in or sign up for Devpost to join the conversation.