All Projects
Own product
2025, AI & Experiments, Bots & Parsers
Documents → structured data
Batch processing of scanned and PDF shipping documents: OCR reads the text, a language model extracts the required fields into JSON.
Concept1 min read
- Input
- PDF, JPG
- Output
- JSON
- Demo
- In progress
Problem
Extract shipment number, container number and forwarder from scanned shipping documents without manual entry.
What was built
- Batch processing of a folder of documents.
- PDF to image conversion.
- Text recognition with Google Vision.
- Field extraction into timestamped JSON with a language model.
Engineering
- A prompt with an example answer and low temperature for a stable format.
