Back to full roadmap
topiccore
Document Understanding (PDF / Tables)
PDF → structured data. Forms, invoices, contracts, reports.
3 hours1 resources1 prereqs
3 approaches:
- Native PDF input (Claude, Gemini) — reads the page visually
- OCR + LLM — Tesseract/Azure OCR → text → LLM
- Layout-aware — Marker, Unstructured, Reducto — preserves table/heading/figure structure
For table-heavy docs, option 3 is most reliable.