Document-extraction SaaS
A SaaS platform that turns raw documents (PDFs, scans, emails) into structured, verifiable data. Designed, operated and maintained in-house under Updev Solutions, for French companies handling high document volumes.
Tech stack
- Python
- FastAPI
- React
- PostgreSQL
- pgvector
- Docling / OCR
- Claude
- Celery
- Docker
- Scaleway (FR)
Framing
Identify document types, fields to extract and business rules. Define confidence thresholds and the acceptable level of human review.
Extraction pipeline
OCR + segmentation, then structured LLM extraction against typed schemas. Every field traces back to its source in the document so it can be verified.
Production
Async queue-based processing, French hosting, Zero Data Retention. A dashboard tracks batches, errors and costs.
Operations
Monitor extraction rates and drift, iterate on prompts and schemas, support users in production.
Screenshots
