AutoPolicy , AI-Powered Insurance Claim Evaluator

Login page
Sign up page
ADK processing
Claim result
Claim result

Inspiration

Filing a car insurance claim is often a slow, manual, and frustrating process. Between collecting documents, proving legitimacy, and waiting for approval, the lack of automation makes the customer experience cumbersome. With the rise of autonomous agents and Google Cloud’s Agent Development Kit (ADK), we saw a huge opportunity to bring AI into insurance claim processing. We imagined a system where intelligent agents handle intake, document verification, fraud detection, and policy validation , reducing both effort and error.

What it does

AutoPolicy is an AI-powered web application that simulates a real insurance claim processing system — from login to decision. ✍️ User Authentication with Email Verification Secure sign-up and login using Firebase Auth with mandatory email verification. 📝 Claim Form Submission A dynamic, user-friendly form captures all key insurance claim details, including accident description, vehicle info, and policy dates. 🤖 Multi-Agent Claim Evaluation Using ADK, a pipeline of agents (Intake, Document Verifier, Fraud Detector, Policy Matcher, Damage Estimator) processes each submission. 📦 Document Upload Users can attach supporting files (e.g., police reports, damage photos) via Firebase Storage. ✅ Smart Decision Engine The Decision Summarizer Agent gives an Approval/Denial verdict with full transparency on fraud scores and policy checks. 📄 Claim Summary Dashboard The final view includes a detailed explanation of why the claim was approved or denied.

How we built it

Frontend: React + TypeScript with Tailwind CSS and Lucide Icons. Hosted on Firebase. Backend: FastAPI-based ADK agent pipeline and hosted on Render. Agents: Created using Google’s open-source Agent Development Kit in Python. Firebase: Used for authentication, email verification, document uploads, and storing user details in Firestore. Hosting: Full-stack deployment with CORS support and Firebase hosting.

Challenges we ran into

Making ADK agents compatible with FastAPI async flow.
Preventing auto-login before email verification in Firebase.
Handling form data dynamically while integrating Firebase document uploads.
Designing a UI that is clean yet detailed enough for insurance professionals and claimants.

Accomplishments that we're proud of

Built a fully working end-to-end multi-agent AI system in just a few days.
Designed a beautiful and responsive UI/UX from scratch.
Implemented email verification with Firebase Auth and smooth user onboarding.
Successfully hosted backend (Render) and frontend (Firebase) in production.
Created a production-level claim flow that mimics real insurance workflows.

What we learned

Deep understanding of Google’s Agent Development Kit and how to design custom agents.
Using Firebase for secure authentication and real-time data storage.
Working with FastAPI and React together.
Handling CORS, deployment flows, and secure document handling.
Building a cloud-native solution with scalable architecture.

What's next for AutoPolicy , AI-Powered Insurance Claim Evaluator

☁️ Cloud-Backed Document Upload: Currently using Firebase Storage; future versions can integrate with Google Cloud Storage for secure, scalable, enterprise-grade file handling. 📊 Claim History Dashboard: Users will be able to view, download, and manage all past claims from their profile. 🧠 Smarter Fraud Detection: Integrate real-world insurance datasets and tools like Vertex AI to power better fraud scoring models. 📱 Mobile-Friendly PWA Version: Convert AutoPolicy into a Progressive Web App for easy mobile access, especially in low-connectivity regions. 📨 Notifications & Alerts: SMS and email notifications (via SendGrid, Twilio, or Firebase Cloud Messaging) for status updates. Can be extended to othe forms of insurance.