A structured-data PDF platform: design a form visually, generate a normal-looking PDF and extract its data back out as clean structured output instantly and with 100% accuracy.
Live demo: https://veraxis-fe.netlify.app
Veraxis has two parts:
Builder - a drag-and-drop canvas where you place fields (Name, Email, Phone, Date, Number, Notes) onto an A4 or Letter page. Type in values and generate a real PDF.
Extractor - upload any PDF created by Veraxis and the data is pulled out instantly. View it as plain text, JSON, or a table, and download as TXT, JSON, Excel or Word.
The PDF that comes out of the Builder looks like a normal document to anyone who opens it - but it carries its data with it, ready to be read back out.
Most "smart" PDF tools (Dext, AutoEntry, Lightyear, etc.) work by scanning a finished document and guessing what the data means. That's OCR - optical character recognition. It's genuinely impressive technology, but it's solving the wrong problem: the data was typed into a computer in the first place, then printed/exported to PDF, then scanned and re-guessed back into data. Every guess has a margin of error (typically 95–99%), which means someone still has to check and fix the output.
Veraxis skips the guessing entirely. When a PDF is generated, the structured values are embedded directly into the file as hidden, read-only form fields alongside the visible layout. Extraction doesn't read pixels - it reads the data that was always there. The result is 100% accurate, every time, because there's nothing to interpret.
The trade-off: this only works for PDFs generated by Veraxis. It can't retrofit arbitrary existing PDFs the way OCR can. That's the deliberate bet - accuracy and reliability for documents created going forward, rather than imperfect best-effort on documents from the past.
Frontend
- React + Vite
@dnd-kit/corefor drag-and-drop field placementxlsx(SheetJS) for Excel exportdocxfor Word export- Custom glass-morphism UI (no component library) - backdrop blur, layered translucency, ambient gradients
Why hidden form fields instead of a metadata-only approach? PDF form fields are a native, standards-based way to embed structured data that survives copying, re-saving, and most PDF tooling - without relying on custom metadata blocks that other software might strip.
Why per-field input validation (Name, Email, Phone, Date, Number, Notes)? Each field type constrains and formats input appropriately on the canvas (e.g. names allow accented international characters, phone fields accept any country format, notes are capped and word-wrapped to fit exactly within their rendered space on the PDF). This keeps the generated document predictable and the extracted data clean.
Why a single-page MVP? Multi-page support and templates were deliberately deferred until there was a real use case to design around, rather than building speculative complexity. The architecture (fields as a flat array with x/y coordinates) was kept simple on purpose so it can be restructured cleanly into a page-based model if/when that's needed.
Why a "Created with Veraxis" watermark? Standard freemium pattern - free-tier output is clearly branded, with the watermark intended to be removable on a paid tier.
This was built as a focused MVP / portfolio project: end-to-end working product, deployed and live, covering the full loop from form design → PDF generation → data extraction → multi-format export.