Pixel-Perfect HTML from Azure Form Recognizer

I need a Python service that takes Azure Document Intelligence (Form Recognizer) JSON from PDFs or images and returns a 1:1 pixel-perfect HTML reconstruction of each page (no raster overlays). This service should be exposed via a small REST API for our Node.js CRM to submit jobs and receive the HTML.

Requirements:
- Exact replication of:
- Text and fonts
- Layouts and structures
- Images and graphics
- Handle primarily scanned images with text

Ideal Skills and Experience:
- Proficiency in Python
- Experience with Azure Document Intelligence
- Strong knowledge of HTML and REST APIs
- Familiarity with Node.js is a plus
- Attention to detail for pixel-perfect accuracy

Please provide examples of similar work done.