← All work

BustPDF

Earlier PDF, OCR, and structured-extraction work worth keeping as reusable document infrastructure.

Why I built it

Complex documents require dependable extraction before useful search, explanation, or learning can occur.

How it works

  1. Document
  2. Extract
  3. Structure
  4. Reuse

Document ingestion feeds text, table, image, and OCR extraction before downstream learning experiments.

Where it stands today

A broad demonstration existed and is offline now; original maintainable extraction components still need to be separated from experiments.

Working now

  • PDF parsing
  • OCR integration
  • Structured extraction experiments
  • AI summary, quiz, and voice experiments over extracted text

Not there yet

  • Provenance, dependency, and maintainability audit remains open.

How it got here

  1. First version: PDF text, table, image, and OCR extraction
  2. Moved to FastAPI and exposed through MCP
  3. Next.js front end with AI summaries, quizzes, and voice chat about the document; deployed

What comes next

Next step

Inventory original parsing and OCR components.

Later ideas, not current features

  1. Extract one tested document-infrastructure component

What I checked, and what stays bounded

Dated evidence

  1. Earlier broad demonstration deployeddeployment

Boundaries it keeps

  • Not a current broad PDF product
  • Only original maintainable components are release candidates

My role

Implemented and integrated multiple extraction and document-interaction paths.