Skip to content

Mehedi26696/DenaAI

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
backend
backend
 
 
frontend
frontend
 
 
.gitignore
.gitignore
 
 
README.md
README.md
 
 

Repository files navigation

🏦 DenaAI - Intelligent Loan Underwriting System

DenaAI Banner FastAPI Flutter XGBoost License

AI-Powered Loan Decision System with OCR, LLM Extraction, and Explainable ML

Features • Architecture • Installation • API Docs • Demo

📋 Table of Contents

🌟 Overview

DenaAI is an intelligent loan underwriting system that combines cutting-edge AI technologies to automate and enhance the loan approval process. Built for financial institutions, it provides:

  • 📄 Automated Document Processing via OCR
  • 🤖 AI-Powered Data Extraction using Groq LLM
  • 🎯 ML-Based Risk Assessment with XGBoost
  • 🔍 Explainable AI for transparent decision-making
  • 📊 Bias Detection & Monitoring for fair lending
  • 👥 Dual Interface for customers and bank officers

Why DenaAI?

Traditional loan processing is slow, subjective, and prone to bias. DenaAI solves these problems by:

✅ Reducing processing time from days to minutes
✅ Providing transparent, explainable decisions
✅ Detecting and preventing algorithmic bias
✅ Supporting both automated and manual review workflows
✅ Helping rejected applicants improve their applications

✨ Features

🎯 Core Features

For Loan Applicants

  • 📱 Mobile-First Application - Apply for loans from anywhere
  • 📸 Smart Document Upload - OCR automatically extracts data from income documents
  • ⚡ Instant Decisions - Get approval/rejection in seconds
  • 💡 AI-Powered Guidance - Understand why you were rejected and how to improve
  • 📊 Application Tracking - Monitor your loan status in real-time
  • 📜 Loan History - View all past applications and outcomes

For Bank Officers

  • 🔍 Manual Review Queue - Handle edge cases requiring human judgment
  • 📊 Comprehensive Dashboard - Monitor all applications
  • 👥 User Management - Manage customer accounts
  • 📈 Analytics & Reports - Track approval rates and trends
  • ✅ Approval Workflow - Approve, reject, or request additional information

For Administrators

  • 🖥️ Web Monitoring Dashboard - Real-time system monitoring
  • 🎯 Bias Detection - Identify and mitigate algorithmic bias
  • 📋 Audit Logs - Complete audit trail of all decisions
  • 📊 Advanced Analytics - Deep insights with interactive graphs
  • ⚙️ System Configuration - Manage thresholds and rules

🔐 Security & Compliance

  • 🔒 Secure authentication for users and admins
  • 📝 Complete audit trail for regulatory compliance
  • 🛡️ Data privacy protection
  • ⚖️ Fair lending compliance monitoring

🏗️ Architecture

┌─────────────────────────────────────────────────────────────┐
│                       USER INTERACTION                       │
│  📱 Flutter Mobile App     🖥️ Flutter Web Dashboard         │
└────────────────┬──────────────────────────────┬─────────────┘
                 │                              │
                 ▼                              ▼
┌─────────────────────────────────────────────────────────────┐
│                    FASTAPI BACKEND                          │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐     │
│  │   Auth API   │  │  Loan API    │  │  Admin API   │     │
│  └──────┬───────┘  └──────┬───────┘  └──────┬───────┘     │
└─────────┼──────────────────┼──────────────────┼─────────────┘
          │                  │                  │
          ▼                  ▼                  ▼
┌─────────────────────────────────────────────────────────────┐
│                    PROCESSING LAYER                          │
│  ┌─────────┐  ┌──────────┐  ┌─────────┐  ┌──────────┐     │
│  │   OCR   │→ │ Groq LLM │→ │  Rules  │→ │ XGBoost  │     │
│  │ Service │  │Extraction│  │  Engine │  │   ML     │     │
│  └─────────┘  └──────────┘  └─────────┘  └──────────┘     │
└─────────────────────────────────────────────────────────────┘
          │                                         │
          ▼                                         ▼
    Decision Result                        Explanation + Score

Workflow

graph TD
    A[User Uploads Document] --> B{User Type?}
    B -->|New User| C[OCR Extraction]
    B -->|Existing User| D[Fetch History]
    C --> E[Groq LLM Parse Fields]
    D --> E
    E --> F[Rule-Based Checks]
    F --> G[XGBoost Prediction]
    G --> H{Decision}
    H -->|Approved| I[Show Explanation]
    H -->|Rejected| J[AI Help & Suggestions]
    H -->|Uncertain| K[Manual Review Queue]
    K --> L[Bank Officer Review]
    I --> M[Store Audit Log]
    J --> M
    L --> M
Loading

🛠️ Technology Stack

Backend

  • Framework: FastAPI 0.124.4
  • ML Engine: XGBoost 3.1.2, Scikit-learn 1.6.1
  • LLM: Groq (Llama 3.1-8B-Instant)
  • OCR: OCR.space API
  • Language: Python 3.10+

Frontend

  • Framework: Flutter 3.0+
  • State Management: Provider/Riverpod
  • Platforms: iOS, Android, Web

Machine Learning

  • Algorithm: XGBoost Gradient Boosting
  • Features: 11 financial indicators
  • Training Dataset: loan_approval_dataset.csv
  • Model Format: Joblib pipeline

DevOps & Tools

  • API Documentation: Swagger UI (FastAPI auto-generated)
  • Environment Management: Python dotenv
  • Version Control: Git

📦 Installation

Prerequisites

  • Python 3.10+
  • Flutter SDK 3.0+
  • Git

Backend Setup

# Clone the repository
git clone https://github.com/Mehedi26696/Fintech.git
cd DenaAI/backend

# Create virtual environment
python -m venv venv

# Activate virtual environment
# Windows
venv\Scripts\activate
# Linux/Mac
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Create .env file
copy .env.example .env  # Windows
cp .env.example .env    # Linux/Mac

# Edit .env and add your API keys
# GROQ_API_KEY=your_groq_api_key
# OCR_SPACE_API_KEY=your_ocr_api_key

# Run the server
uvicorn src.main:app --reload --host 0.0.0.0 --port 8000

Frontend Setup

cd ../frontend

# Install dependencies
flutter pub get

# Run on desired platform
flutter run -d chrome        # Web
flutter run -d windows       # Windows
flutter run                  # Mobile (connect device/emulator)

🚀 Usage

Starting the Backend

cd backend
uvicorn src.main:app --reload --port 8000

Access the API documentation at: http://localhost:8000/docs

Running the Frontend

cd frontend
flutter run

API Health Check

curl http://localhost:8000/health

Response:

{
  "status": "healthy",
  "service": "Loan Underwriting & Prediction System",
  "version": "2.0.0"
}

📚 API Documentation

Base URL

http://localhost:8000

Key Endpoints

1. Document-Based Underwriting

POST /underwrite
Content-Type: multipart/form-data

Parameters:
- file: Document image/PDF (required)
- loan_amount: float (optional)
- loan_term: int (optional)
- cibil_score: int (optional)
- income_annum: float (optional)
- ... (other overrides)

Response:
{
  "extracted": {...},
  "rule_checks": {...},
  "approval_probability": 0.78,
  "decision": "Approved",
  "explanation": "...",
  "missing_fields": []
}

2. Feature-Based Prediction

POST /model/predict

Body:
{
  "features": {
    "no_of_dependents": 2,
    "income_annum": 600000,
    "loan_amount": 500000,
    "loan_term": 12,
    "cibil_score": 750,
    ...
  },
  "threshold": 0.5
}

Response:
{
  "approval_probability": 0.82,
  "decision": "Approved",
  "threshold": 0.5
}

3. Batch Prediction

POST /model/predict_batch

Body:
{
  "records": [
    {"features": {...}},
    {"features": {...}}
  ],
  "threshold": 0.5
}

4. Model Schema

GET /model/schema

Response: Complete feature schema with descriptions

Authentication Endpoints (To be implemented)

POST /auth/login
POST /auth/signup
POST /auth/forgot-password

📱 Frontend Architecture

frontend/lib/
├── core/                      # Core configurations
│   ├── config/               # App & API config
│   ├── theme/                # App theme & colors
│   ├── routes/               # Navigation
│   ├── constants/            # Constants
│   └── utils/                # Utilities
│
├── features/                  # Feature modules
│   ├── splash/               # Splash screen
│   ├── onboarding/           # Onboarding flow
│   ├── auth/                 # Authentication
│   │   ├── screens/          # Login, signup, etc.
│   │   ├── widgets/          # Auth widgets
│   │   ├── services/         # Auth service
│   │   └── models/           # User models
│   │
│   ├── user/                 # User dashboard
│   │   ├── screens/          # Home, profile, settings
│   │   └── widgets/          # Dashboard widgets
│   │
│   ├── loan/                 # Loan management
│   │   ├── screens/          # Application, upload, status
│   │   │   ├── loan_form_screen.dart
│   │   │   ├── document_upload_screen.dart
│   │   │   ├── loan_explanation_screen.dart
│   │   │   ├── ai_help_screen.dart
│   │   │   └── loan_history_screen.dart
│   │   ├── widgets/          # Loan widgets
│   │   ├── services/         # Loan API services
│   │   └── models/           # Loan models
│   │
│   ├── admin/                # Admin (mobile/tablet)
│   │   ├── screens/          # Management screens
│   │   │   ├── manual_review_screen.dart
│   │   │   └── review_detail_screen.dart
│   │   ├── widgets/          # Admin widgets
│   │   ├── services/         # Admin services
│   │   └── models/           # Admin models
│   │
│   └── admin_web/            # Web-only monitoring
│       ├── screens/          # Web dashboards
│       │   ├── admin_dashboard.dart
│       │   ├── bias_monitor.dart
│       │   ├── audit_log.dart
│       │   └── graphs.dart
│       ├── widgets/          # Web widgets
│       ├── services/         # Monitoring services
│       └── models/           # Monitoring models
│
└── shared/                    # Shared components
    ├── widgets/              # Reusable widgets
    ├── services/             # HTTP, storage, etc.
    ├── models/               # Common models
    ├── exceptions/           # Error handling
    └── extensions/           # Dart extensions

🤖 Machine Learning Pipeline

Model Training

cd backend
jupyter notebook "Training Code.ipynb"

Features Used (11 total)

  1. no_of_dependents - Number of dependents
  2. income_annum - Annual income
  3. loan_amount - Requested loan amount
  4. loan_term - Loan term in months
  5. cibil_score - Credit score (300-900)
  6. residential_assets_value - Residential property value
  7. commercial_assets_value - Commercial property value
  8. luxury_assets_value - Luxury assets value
  9. bank_asset_value - Bank deposits/savings
  10. education - Education level (encoded)
  11. self_employed - Employment type (binary)

Model Performance

  • Algorithm: XGBoost with hyperparameter tuning
  • Training Accuracy: ~95%
  • Validation Accuracy: ~92%
  • F1 Score: ~0.91
  • Model File: loan_xgb_pipeline.joblib

Rule Engine

Business rules applied before ML prediction:

  • ✅ Income reasonability check
  • ✅ CIBIL score minimum threshold
  • ✅ Loan-to-income ratio validation
  • ✅ Missing field detection
  • ✅ OCR confidence validation

🎨 Key Features Breakdown

1. Explainable AI

Every decision comes with:

  • Probability score
  • Rule check results
  • Feature importance
  • Clear explanation text

2. Bias Detection

Web dashboard monitors:

  • Approval rate by demographics
  • Score distribution fairness
  • Statistical parity checks
  • Disparate impact analysis

3. AI-Powered Help

Rejected applicants receive:

  • Reasons for rejection
  • Actionable improvement suggestions
  • What-if scenario analysis
  • Personalized guidance via Groq LLM

4. Multi-Level Decision Flow

Automatic Approval → User gets instant result
Automatic Rejection → AI help + explanation
Uncertain Cases → Manual review by officer

🔒 Environment Variables

Create a .env file in the backend/ directory:

# API Configuration
API_TITLE=Loan Underwriting & Prediction System
API_VERSION=2.0.0
DEBUG=False

# ML Model
MODEL_PATH=loan_xgb_pipeline.joblib
DEFAULT_APPROVAL_THRESHOLD=0.5

# OCR Service
OCR_SPACE_API_KEY=your_ocr_api_key_here
OCR_SPACE_BASE_URL=https://api.ocr.space/parse/image

# Groq API
GROQ_API_KEY=your_groq_api_key_here
GROQ_BASE_URL=https://api.groq.com/openai/v1/chat/completions
GROQ_MODEL=llama-3.1-8b-instant

# Optional: Gemini API (backup)
GEMINI_API_KEY=your_gemini_api_key_here

Getting API Keys

🤝 Contributing

We welcome contributions! Please follow these steps:

  1. Fork the repository
  2. Create a feature branch (git checkout -b feature/AmazingFeature)
  3. Commit your changes (git commit -m 'Add some AmazingFeature')
  4. Push to the branch (git push origin feature/AmazingFeature)
  5. Open a Pull Request

Development Guidelines

  • Follow PEP 8 for Python code
  • Follow Flutter/Dart style guide
  • Write tests for new features
  • Update documentation

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

👥 Team

DenaAI Development Team

H.M. Mehedi Hasan
H.M. Mehedi Hasan
Backend & ML Development 		Chowdhury Shafahid Rahman
Chowdhury Shafahid Rahman
API Integration 		Sumaiya Rahman Soma
Sumaiya Rahman Soma
UX/UI & Bias Detection 		MD. Abu Bakar Siddique
MD. Abu Bakar Siddique
Frontend Development

🙏 Acknowledgments

  • FastAPI for the excellent web framework
  • XGBoost team for the ML library
  • Groq for LLM API access
  • OCR.space for document processing
  • Flutter team for the cross-platform framework

📞 Support

For support, email support@denai.com or open an issue on GitHub.

Built with ❤️ by the DenaAI Team

GitHub Stars GitHub Forks

About

DenaAI is an intelligent loan underwriting system that combines cutting-edge AI technologies to automate and enhance the loan approval process.

Resources

Readme
Activity

Stars

0 stars

Watchers

1 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages