🛡️ InsureTech 360: Intelligent Claims & Risk Assessment System

📌 Project Overview

InsureTech 360 is an industry-level, full-stack AI application designed as an Internal Agent-Assist Dashboard to modernize the vehicle insurance claim process. Built specifically for Customer Support teams and Claims Adjusters, this platform reduces Average Handling Time (AHT) by integrating multiple AI disciplines to process "messy" real-world data instantly.

It unifies Computer Vision (CNN) for rapid damage assessment, Deep Learning (ANN) for fraud detection, and Generative AI (RAG) for real-time policy assistance into a single, unified workspace.

---

🚀 Key Features

• 📸 Visual Inspection (CNN): Automatically detects car damage types and severity from uploaded images using deep Convolutional Neural Networks.
• 🔍 Fraud Detection (ANN): Analyzes tabular user data (policy details, history) to predict the probability of a claim being fraudulent using Artificial Neural Networks.
• 🤖 Smart Policy Assistant (RAG): A GenAI-powered chatbot (built with LangChain & Vector DB) that answers user queries based on specific policy PDF documents.
• 📊 Dynamic Dashboard: A professional UI built with HTML/CSS and Flask to visualize risk scores and analysis reports.
• 🏭 Modular Architecture: Follows industry-standard coding practices (Pipelines, Components, Logging, Exception Handling) for scalability.

---

🛠️ Tech Stack

• Languages: Python 3.10
• Backend Framework: Flask
• Machine Learning: Scikit-learn, Pandas, NumPy
• Deep Learning: TensorFlow/Keras (CNN, ANN)
• Generative AI: LangChain, FAISS (Vector DB), OpenAI/Google Gemini API
• Frontend: HTML5, CSS3, Bootstrap 5
• DevTools: Git, Dotenv

---

📂 Datasets & Resources Used

To run this project locally, you will need to download the following datasets and place them in the artifacts/data/ folder as per the structure mentioned below.

1. Insurance Fraud Detection Dataset (CSV)

This dataset is used to train the ANN model for predicting fraudulent claims.

• Source: Kaggle (Vehicle Claim Fraud Detection)
• Download Link: Click Here to Download CSV
• Placement: Extract and rename the file to insurance_claims.csv inside artifacts/data/.

2. Car Damage Detection Dataset (Images)

This dataset is used to train the CNN model to detect if a car is damaged or whole.

• Source: Kaggle (Car Damage Detection)
• Download Link: Click Here to Download Images
• Placement: Extract the folders (00-damage, 01-whole) inside artifacts/data/car_images/.
  • Ensure structure: artifacts/data/car_images/data/training/ and validation/.

3. Policy Document for Chatbot (RAG)

This PDF is used by the Gemini-powered chatbot to answer user queries regarding insurance policies.

• Sample File: You can use any standard Car Insurance Policy PDF.
• Example Link: Sample Policy PDF (Generic)
• Placement: Save the file as policy.pdf inside artifacts/data/.

---

📂 Project Structure

The project follows a modular, production-ready structure:

InsureTech_360/
├── artifacts/                      # Stored Models, Preprocessors & Outputs (Ignored in Git)
├── logs/                           # Application & Pipeline Logs
├── notebooks/                      # Jupyter Notebooks for EDA & Experiments
├── src/                            # Source Code
│   ├── components/                 # Core ML Components
│   │   ├── data_ingestion.py       # Data Collection & Loading
│   │   ├── data_transformation.py  # Feature Engineering & Preprocessing
│   │   ├── model_trainer.py        # Model Training Logic
│   │   └── model_evaluation.py     # Model Evaluation & Metrics
│   ├── pipelines/                  # ML Pipelines
│   │   ├── training_pipeline.py    # End-to-End Training Pipeline
│   │   └── prediction_pipeline.py  # Inference / Prediction Pipeline
│   ├── utils/                      # Helper & Utility Functions
│   │   └── common.py
│   ├── logger.py                   # Custom Logging Configuration
│   ├── exception.py                # Custom Exception Handling
│   └── constants.py                # Project-wide Constants & Configs
├── static/                         # CSS, JS, Images (Frontend Assets)
├── templates/                      # HTML Templates (Frontend)
├── .env                            # Environment Variables (Ignored in Git)
├── app.py                          # Application / API Entry Point
├── requirements.txt                # Project Dependencies
├── setup.py                        # Package Setup Configuration
└── README.md                       # Project Documentation

---

⚙️ Installation & Setup

1. Clone the Repository

   git clone https://github.com/kunalbaghelkb/InsureTech_360.git && cd InsureTech_360
   

2. Create Virtual Environment

   python3.11 -m venv .venv
   
   # Windows
   .venv\Scripts\activate
   
   # Mac/Linux
   source .venv/bin/activate
   

3. Install Dependencies

   pip install -r requirements.txt
   

4. Set Environment Variables Create a .env file in the root directory and add your keys

   GEMINI_API_KEY=your_api_key_here
   GOOGLE_LLM_MODEL_NAME=gemini-2.5-flash (configure as needed)
   GOOGLE_EMBED_MODEL_NAME=gemini-embedding-001 (configure as needed)
   

5. Generate Model Execute this file to generate the model (Ensure that all required datasets and resources have been added as specified.)

   python src/pipelines/training_pipeline.py
   

6. Run the Application

   python app.py
   

---

👨‍💻 Author

Kunal Baghel

Aspiring Data Scientist & AI Engineer

LinkedIn | GitHub