AI Branch Manager – Video-Based Loan Assistance

The AI Branch Manager is a web-based platform designed to provide a digital, branch-like experience for users applying for loans. Using AI-powered video interactions, document processing, and rule-based eligibility checks, this system guides users through the loan application process in an intuitive and interactive manner.

A demonstration video showcasing the AI Branch Manager's functionality is available.

This demo highlights the key features and user flow of the application, providing a visual overview of how the system works.

• Users interact with a pre-recorded AI video assistant that mimics a real-life bank manager.
• The AI Manager asks structured financial questions and provides step-by-step guidance on loan applications.

• Users respond to financial queries by recording video responses.
• Basic facial verification ensures that the same applicant continues throughout the process, preventing fraud.

• Users upload images of Aadhaar, PAN, income proof, address proof, and bank statements via mobile or webcam.
• AI-powered OCR extracts key details such as:
  • ID Documents: Name, DOB, ID number, address
  • Income Documents: Monthly income, employer name, pay period
  • Address Proof: Full address, pincode, city, state
  • Bank Statements: Account number, bank name, balance, IFSC code

• A rule-based system evaluates loan eligibility using user responses and document data.
• Instant feedback provided:
  • ✅ Approved
  • ❌ Rejected (with reasons)
  • 🔄 More Info Needed

• Pre-recorded assistant videos can be available in multiple Indian languages for a better user experience.
• OCR supports multiple languages including English and several Indian languages.

manager/
│── CapitalCue - Your AI Branch Manager.mp4  # Demo video in root directory
│── processing/                 # Loan application processing logic
│   └── ocr_extraction.py       # Document OCR and data extraction
│── public/                     # Public assets (images, icons, videos)
│── src/
│   ├── app/                    # Main application logic
│   ├── components/             # Reusable UI components
│   ├── pages/
│   │   ├── api/                # Backend API routes
│   │   ├── index.tsx           # Landing page
│   ├── utils/                  # Utility functions
│── .gitignore                  # Files to be ignored by Git
│── README.md                   # Project documentation
│── requirements.txt            # Python dependencies
│── next-env.d.ts               # TypeScript environment definitions
│── next.config.ts              # Next.js configuration
│── package.json                # Project dependencies and scripts
│── postcss.config.mjs          # PostCSS configuration
│── tsconfig.json               # TypeScript configuration

Ensure you have the following installed:

• Node.js (>= 16.x)
• Python (>= 3.7)
• Git
• Yarn or npm
• Tesseract OCR (for document processing)

1. Clone the repository

   git clone https://github.com/sujallimje/manager.git
   cd manager

2. Install JavaScript dependencies

   npm install  # or yarn install

3. Install Python dependencies

   pip install -r requirements.txt

4. Install Tesseract OCR

   • Windows: Download from GitHub
   • macOS: brew install tesseract
   • Linux: sudo apt install tesseract-ocr

   Update the path in processing/ocr_extraction.py if needed:

   pytesseract.pytesseract.tesseract_cmd = r'path/to/tesseract.exe'

5. Run the development server

   npm run dev  # or yarn dev

6. Change firebase.js and groq key

   • src/lib/firebase.js
   • src/app/api/chat/route.ts

7. Access the application

   • Open http://localhost:3000 in your browser.
   • To view the demo video, access the CapitalCue - Your AI Branch Manager.mp4 file in the root project directory.

The system uses advanced OCR techniques to extract information from various documents:

1. ID Documents (Aadhaar & PAN):

   • Automatically detects document type
   • Extracts name, DOB, ID number, address, gender
   • Supports multiple languages

2. Income Documents:

   • Extracts monthly income, employer name, employee ID
   • Identifies pay period

3. Address Proof:

   • Extracts full address, pincode, city, state

4. Bank Statements:

   • Extracts bank name, account number, IFSC code, balance
   • Identifies statement period

• Frontend: Next.js, React.js, Tailwind CSS
• Backend: Next.js API Routes, Node.js
• AI Processing:
  • OpenCV (for image preprocessing and facial verification)
  • Tesseract OCR (for document text extraction)
  • pytesseract (Python wrapper for Tesseract)
  • NumPy (for image manipulation)
• Database: MongoDB (optional, for storing loan applications)
• Deployment: Vercel

• Implement AI-powered chat/video assistant using NLP.
• Add multilingual support with AI-generated voice responses.
• Integrate with banking APIs for real-time loan approval processing.
• Enhance OCR accuracy with deep learning-based document analysis.

• Add a pull request