Research & models / 2024—26
Document intelligence.
Structured extraction work for Kenyan identity and onboarding documents.
Research & modelsDocument intelligenceSystems shaped for people.
01 / Context
The problem behind the work.
Applications slow down when document fields must be re-keyed or cannot be checked consistently.
02 / Contribution
What I built.
- 01
FastAPI extraction paths combining image/PDF processing, OCR, Gemini vision and rule-based validation.
- 02
Structured JSON outputs for document types including IDs and registration records.
03 / Where it stands
This workstream is separate from the Django/Paperless enterprise document workflow. The linked repositories are public source examples.
Technology and methods
FastAPIPaddleOCROpenCVPyMuPDFGeminiPython