🛡️ InsureGuard: AI-Powered Fraud Insurance Claim Detection Engine

• InsureGuard is a insurance claim analysis platform designed to detect medical insurance fraud in real-time.

• By combining a Deterministic Engine with the reasoning capabilities of Generative AI, it helps auditors identify the inconsistencies instantly.

Live Project Link: https://insureguard-kbut.onrender.com/

👨‍💼 Targeted User: The Claims Auditor

Unlike patient-facing portals, InsureGuard is a specialized internal tool designed for Insurance Claims Officers and Analysts.

The Workflow:

1. Physical Review: The analyst receives a formal insurance claim application (the hard copy or digital PDF).
2. Data Extraction: The analyst identifies key parameters—such as diagnosis codes, specific treatments, hospitalization dates, and exact billing figures.
3. System Entry: These verified parameters are entered into the InsureGuard portal.
4. Instant Audit: The system processes the manual input to provide an immediate risk assessment, allowing the officer to approve genuine claims faster or investigate flagged frauds with AI-backed reasoning.

⚖️ Strategic Design: Why Deterministic Engine for Prediction?

A key architectural decision in InsureGuard is the separation of Prediction and Explanation:

• Non-Generative Prediction: We do not use Generative AI (LLMs) to determine the fraud risk score or decide the claim's status.
• Accuracy & Reliability: Generative AI can occasionally suffer from "hallucinations" or inconsistent reasoning, which is unacceptable when dealing with financial payouts and people's health coverage.
• Fairness for Genuine Claimants: To protect honest users from being wrongly flagged by a "black box" algorithm, all fraud detection is handled by a transparent, Deterministic Engine.
• The Role of Generative AI: Generative AI is used strictly for Audit Assistance—it interprets the already detected values and uses the input parameters to provide a human-readable, point-by-point clinical explanation stating why an application is detected as fraud claim or genuine claim for the analyst.

🚀 Features

• Hybrid Risk Scoring: Uses a weighted logic engine to calculate fraud probability along with percentage risk
• Clinical Reasoning: Integrated with Genearative AI to provide point-by-point explanations of discrepancies present in the report.
• Automated Audit: Instantly flags issues like duration-treatment mismatches and financial anomalies.
• Modern Interface: A clean, responsive UI built with Tailwind CSS for high-speed auditing workflows.

🖼️ System Architecture

🛠️ Tech Stack

• Backend: Python / Flask
• AI: Google Gemini API (genai SDK)
• Frontend: HTML5, Tailwind CSS, JavaScript
• Deployment: Render

📸 Application's Screenshots

1. Landing Page:

A high-impact hero section that introduces InsureGuard as an AI-powered intelligence layer for real-time insurance claim auditing.

2. Audit Form: The Analyst's Workspace

Manual Data Entry Portal: A structured interface where claims officers input key parameters—like diagnosis, treatment, and financials—extracted from physical application forms.

3. Analysis Report: The Fraud Risk Overview

• A comprehensive dashboard displaying the fraud probability percentage and final audit status calculated by the deterministic rule engine.
• Rule Violation Breakdown: A dedicated section that only appears when risks are detected, listing specific red flags like clinical mismatches or financial padding.
• Explainable AI Context: A point-by-point clinical justification generated by Gemini 1.5 Flash, providing the human officer with the "why" behind the flagged data.