Azure AI Document Intelligence Demo

I’m putting together a small proof-of-concept that shows off the newest Azure AI Document Intelligence capabilities. The goal is a runnable demo—plus the clean C#/.NET code behind it—that can classify incoming forms, pull out their key values, automatically redact sensitive content, and then split the document into logical sections.

Scope of work
• Documents: We’ll focus on Forms for now, though I’d like the approach to remain flexible enough that invoices or contracts could be added later.
• Extraction: The demo must return every text field, checkbox state, and any table data the form contains.
• Autosplit: After processing, the code should detect the document’s internal sections and break the file accordingly.
• Auto-redact: Before saving or displaying the output, sensitive strings (names, IDs, addresses, etc.) need to be blacked out.
• Technology stack: Azure AI Document Intelligence (formerly Form Recognizer) using the latest REST/SDK, called from a .NET 6+ console or minimal-API project.

Deliverables
- Visual demo (web or desktop) that lets me upload a sample PDF or image, watch the processing, and download the split, redacted parts.
- Well-commented C# source showing every API call, model reference, and redaction step.
- A brief markdown read-me: setup steps for my Azure subscription, keys, environment variables, and how to extend the model to other document types.
- Sample test files and the JSON output the service returns, so I can verify extraction accuracy.

Acceptance criteria
1. Uploading a multi-page form immediately classifies it, extracts all text fields, checkbox values, and tables with 95 %+ accuracy on the supplied samples.
2. Sensitive data is replaced by mask characters or rectangles in both PDF and image outputs.
3. The resulting document is split exactly at each detected section boundary.
4. All code builds and runs locally with a single command and deploys to Azure with clear instructions.

If you have previous Azure Form Recognizer or AI Document Intelligence experience and can turn this around quickly, I’d love to see a short outline of your approach when you bid.