I will build an ai document extraction prototype with evaluation

Build a runnable AI extraction prototype for one agreed input family: text-based PDFs OR product-label images.

Scope: up to 12 fields, ten authorized supplied examples, one agreed API/model provider and one demo interface. PDFs are limited to ten pages per sample; label images to 10 MB each.

You receive the extraction demo, validated JSON structure, comparison against your supplied expected outputs, documented failures/limits, source code and setup instructions. One correction round covers the agreed schema and examples.

Provide a field schema, non-sensitive or synthetic samples with expected outputs, demo language, provider preference and agreement on client-paid API usage. Share access through the permitted contract workflow.

Acceptance: the prototype runs with the documented setup, returns the agreed structure and includes an honest evaluation on the supplied examples. No perfect-accuracy or clinical-validation promise. Production deployment, extra document families and recurring operation need separate scope. Delivery: 21 calendar days after complete inputs.