Case study · German fintech
Document AI
OCR and AI extraction that reads business documents and hands the important fields on as structured data, with a person as the fallback.
- Role
- Technical lead. Processing pipeline and integrations.
- When
- 2022 – 2024
- Client
- German fintech
Illustration with simulated data. No client data or internals are shown.
The problem
Businesses and their advisers exchange a steady flow of paperwork: invoices, receipts, bank statements, transfer forms, identity documents, register extracts, delivery notes, pension and insurance documents. Someone had to read each one and type the important parts into the right place.
What I built
A document-processing platform that takes documents from customers, email and partner systems, reads them with OCR and AI extraction, and passes the important fields on as structured data. Documents it can’t read reliably go to a person instead of being guessed at.
I also built the integrations that gather documents from external provider APIs, websites and email, so they all arrive in one place.
Outcome
The platform serves more than 1.3 million users across real estate, IT and tourism, and turns manual data entry into review.
My role
Technical lead on the processing pipeline and the document integrations, working with the client’s team in Germany.