Inspiration
The idea for this project came from my internship at an insurance company, where I was exposed to how insurance claims are processed and investigated. I observed that fraud detection is especially difficult when the same incident is claimed across multiple insurance companies. Because insurers operate in isolated systems, there is no efficient way to identify duplicate or repeated claims, which often leads to financial losses and delayed investigations. This real-world challenge inspired me to build a system that could assist insurers in detecting fraud collaboratively without compromising sensitive data.
What it does
Cross-Insure Insurance Company Fraud Detector is a platform that analyzes insurance claims to identify potential fraud across multiple insurers. It allows insurers to submit claim details and supporting images for AI-based analysis. The system detects patterns such as reused images, inconsistent claim details, and similar incidents occurring over time, helping insurers flag suspicious claims early.
How I built it
The application is built using FastAPI for the backend and integrates the Gemini 3 Flash API as its core intelligence engine. Gemini’s multimodal capabilities are used to analyze both structured claim data and uploaded incident images. Images are securely stored in blob storage, while metadata such as incident type, location, severity, and timestamps are analyzed to detect anomalies and recurring fraud patterns.
Challenges we ran into
One major challenge was ensuring data privacy while still enabling cross-insurance fraud detection. Another challenge was deployment and dependency compatibility under tight time constraints, especially when working with AI and image-processing libraries.
Accomplishments that im proud of
I successfully built a working AI-powered fraud detection system inspired by a real industry problem. The platform demonstrates how multimodal AI can be applied to insurance fraud detection while respecting data isolation and privacy requirements.
What I learned
This project deepened my understanding of insurance systems, AI-driven fraud detection and multimodal analysis. I also gained hands-on experience deploying production-ready backend services and integrating advanced AI APIs into real-world applications.
What's next for Cross-Insure Insurance Company Fraud Detector
Future improvements include onboarding multiple insurers, enhancing fraud scoring accuracy, adding dashboards for investigators, and expanding the AI model to analyze documents such as claim forms and reports alongside images. Also find a way to penetrate into the medical aid industry
Log in or sign up for Devpost to join the conversation.