Co-founder project · Document AI
Project Parse case study
Clean JSON from invoices, with a Gemini fine-tuning loop.
Solution · Document AI
For logistics, construction and finance teams retyping data from PDFs.
How it works
Documents
PDFs in any layout
Extraction
Vision and language models
Review
Uncertain fields checked
Retraining
Corrections improve the model
Your systems
JSON, ERP or database
Capabilities
Shipped work
Built by Sortup or one of our co-founders, as labelled.
Co-founder project · Document AI
Clean JSON from invoices, with a Gemini fine-tuning loop.
Co-founder project · Construction
A model per vendor, fine-tuned from examples, then used on new PDFs.
FAQ
A sample of your real documents is enough to start. Corrections made during review become training data.
Yes. Where layouts differ a lot, we fine-tune a model per vendor.
It stays in your accounts wherever possible, and model providers are used under API terms that exclude training on it.
No pitch deck. An engineer, and a straight answer.