17 / Projects
Technical Knowledge Engine
Documents become a knowledge base a model can actually use.
The project is shown at the level of problem, architecture and capabilities; some implementation details remain reserved.
01 · The problem
An LLM on top of unprocessed documents answers fluently and wrongly. Knowledge has to be extracted before it is retrieved.
02 · The idea
Documents → extraction → normalisation → database → knowledge → LLM.
03 · The system
The shared spine between the PDF pipeline, yacht pricing and the engineering assistant.
04 · The implementation
In development as a reusable layer, not a one-off script.
05 · The next step
Make retrieval boring: the right record, with a source, every time.
The architecture
- 01 Documents
- 02 Data extraction
- 03 Normalisation
- 04 Database
- 05 Knowledge
- 06 LLM
Technologies
- Python
- PostgreSQL
- SQLAlchemy
- vLLM
- FastAPI
Categories
Data · AI · Engineering