Azure Document Intelligence – AI-Based PDF Classification & Splitting System

This repository demonstrates a real-world AI document processing system built using Azure Document Intelligence, designed to automatically:

• Classify multi-document PDFs
• Split them into individual files
• Upload each document to the correct SharePoint folder

The solution replicates a production-grade back-office automation workflow used in enterprise document pipelines.

---

🚀 What This Project Does

Given a single multi-page PDF containing different document types (e.g. passport + license):

1. Classifies each page using a trained Azure classifier
2. Identifies document type per page
3. Splits the PDF accordingly
4. Uploads each document to SharePoint
5. Organizes files into business folders automatically

---

🧠 Use Cases

This system is ideal for:

• KYC automation
• Identity verification systems
• Loan onboarding workflows
• Back-office document segregation
• Insurance & banking operations
• Compliance document routing

---

📂 Project Structure

Azure-Document-Intelligence/
│
├── azure_doc_intelligence.py
│   → Simple local PDF classification demo
│
├── classify_split_pdf_using_azure_doc_intelligence.py
│   → Full enterprise pipeline with:
│      - Classification
│      - Splitting
│      - SharePoint upload
│
└── README.md (this file)

---

⚙️ Technology Stack

• Azure Document Intelligence
• Azure Custom Classifier
• Python 3.10+
• PyMuPDF (PDF splitting)
• Microsoft Graph API
• Azure AD (MSAL authentication)
• SharePoint Online
• REST APIs

---

🔑 Key Features

• AI-powered document classification
• Automatic PDF splitting
• Multi-document handling
• Zero manual sorting
• Microsoft Graph integration
• SharePoint upload automation
• Enterprise authentication
• Modular processing pipeline

---

🧩 System Architecture

PDF Input
↓
Azure Document Intelligence
↓
AI Classification Model
↓
Page-level Document Detection
↓
PDF Split Engine
↓
Microsoft Graph API
↓
SharePoint Document Library

---

🔐 Security Design

• Azure AD app authentication
• No hardcoded credentials in production
• Token-based Microsoft Graph access
• OAuth2 client credentials flow
• Sample keys only in repository

---

▶️ How to Run

1. Install dependencies

pip install azure-ai-documentintelligence pymupdf msal requests

2. Configure environment

Edit:

DI_ENDPOINT
DI_KEY
CLASSIFIER_ID
CLIENT_ID
CLIENT_SECRET
TENANT_ID
SITE_URL

3. Execute

python classify_split_pdf_using_azure_doc_intelligence.py input.pdf

---

📈 Why This Project Is Valuable

This is not a tutorial demo.

It demonstrates:

• Production-style AI pipelines
• Real enterprise authentication
• Business document automation
• End-to-end system design
• Cloud-native AI integration

The same architecture is used in:

• Banks
• FinTech platforms
• Insurance companies
• Legal tech systems
• Government document processing

---

🧪 Sample Output

Input:

combined_kyc.pdf

Output:

combined_kyc_passport.pdf → SharePoint/Passport  
combined_kyc_license.pdf  → SharePoint/License  

---

👤 Author

Dinakar S AI Engineer | Automation Engineer | Cybersecurity

---

This is exactly how senior engineers document real systems.