07 / Projects
PDF Intelligence Pipeline
From PDF to structured knowledge: extract, parse, validate, normalise, store.
The project is shown at the level of problem, architecture and capabilities; some implementation details remain reserved.
01 · The problem
Technical knowledge is locked in PDFs. Models can read files; they cannot replace a reliable extraction and validation path.
02 · The idea
A deterministic pipeline first, an AI layer second. Invalid data never becomes a quotation or a decision.
03 · The system
PDF → extract → parse → validate → normalise → SQL → knowledge.
04 · The implementation
Connected to the AI Yacht Pricing System and the Technical Knowledge Engine. SQL as the contract between documents and intelligence.
05 · The next step
Harden validation on messy real-world documents and widen the document types the pipeline can trust.
The architecture
- 01 PDF
- 02 Extract
- 03 Parse
- 04 Validate
- 05 Normalise
- 06 SQL
- 07 Knowledge
Technologies
- Python
- PostgreSQL
- SQLite
- SQLAlchemy
- FastAPI
Categories
Data · AI · Engineering