Car Insurance Claim management

The claim adjustment procedure can be lengthy and susceptible to mistakes. A client from the insurance sector recently shared, “Every time an adjuster gets involved, it costs us.” For every claim, adjusters must sift through the client's prior claims and associated guidelines. These records are often dispersed over various systems and formats, complicating the retrieval of pertinent data and delaying the generation of an accurate payment estimate.

Using a vector database, the adjuster can easily request the AI to "display images resembling this collision." The system, powered by Vector Search, then provides photos of vehicular accidents with matching damage patterns from the claims history. This allows the adjuster to swiftly juxtapose the accident photos with the most pertinent ones from the insurer's claim archives.

This application ix trying to showcase the above. STILL WIP

Update config.ini file

Updated URI with your appropriate credentials

[MongoDB]
URI = mongodb+srv://user:password@<cluster_name>.tcrpd.mongodb.net/?retryWrites=true&w=majority
DATABASE = insurance_company
COLLECTION = images

Encode images and store vectors in MongoDB Atlas

Will install python requirements and run python application to encode images located in images folder

pip install git+https://github.com/huggingface/transformers 
pip install -r req.txt
python3.11 app.py --path images              

Create Atlas Search Index

Create Atlas Search index, on database insurance_company and collection images using the JSON config and lappy below config

{
    "mappings": {
        "dynamic": true,
        "fields": {
            "imageEncodings": {
                "dimensions": 128,
                "similarity": "euclidean",
                "type": "knnVector"
            },
            "captionEncodings": {
                "dimensions": 128,
                "similarity": "euclidean",
                "type": "knnVector"
            }            
        }
    }
}

Set up and start the Beckend

Navigate to the frontend directory and install the required npm packages:

python3.11 app.py --server 

Set up and start the Frontend

Navigate to the frontend directory and install the required npm packages:

cd claim-management-frontend
npm install
npm start

The React application should now be running on http://localhost:3000, and the Flask application should be running on http://localhost:5000.

Using the Application

1. Open your web browser and navigate to http://localhost:3000.
2. Click the choose file button to select a picture to upload match against reported claims
3. The application will encode the image and interact with the backend to retrieve related similar claims