See what the local PDF engine really returns
This demonstration is generated from synthetic training PDFs, not from a customer policy or insurer-approved illustration.
What appears in the evidence
The parser records original page, text line, candidate labels, raw number strings and PDF fingerprint. The report deliberately does not certify which table column a number belongs to. A reviewer must validate the year, header and original document before proceeding.
Open the synthetic sample evidence
View sample batch manifest (JSON) Download local Mac preview
What this does not demonstrate
No scanned OCR, no third-party insurer integration, no approval for professional reliance, no real payment or live customer transaction. A claim of accuracy on unseen insurer documents would require independent reference cases and validation.
Independent educational software prototype. Not affiliated with an insurer or regulator. No individual advice, official quotes, coverage, underwriting or transactions are provided. Information reviewed for prototype on 11 Nov 2026.