ProGen

From messy documents to structured data, automatically

Turn any document into usable data.

ProGen reads invoices, receipts, purchase orders, and EPC project documents, and turns them into structured, searchable data your systems can actually use. No templates. No manual entry.

ProGen Extraction

Vendor

Summit Steel & Supply Co.

Invoice #

INV-20482

Amount

$18,420.00

Line itemQtyPrice
Structural steel beams12$980.00
Galvanized fasteners500$0.85
Site delivery1$620.00

Built for teams that process documents at volume

Self-hosted LLMNo per-page vendor lock-inBuilt for EPC, finance & ops teams

See ProGen in 90 seconds.

Upload a document. Watch it get read, structured, and turned into clean data, live.

Try it on a real document.

Drop in a sample invoice or receipt and watch ProGen extract the data in real time.

Waiting for a document

Structured output will appear here

Want this on your own documents? Request Access

90%+

fields extracted automatically, no manual entry

Minutes, not hours

from raw document to structured record

100%

layout and table structure preserved, line items stay mapped to prices

01 · Ingestion

Feed it anything.

Scanned images, multi-page PDFs, or unstructured files, ProGen ingests it all in one pipeline, no pre-sorting required.

  • Images and multi-page PDFs in one upload
  • Any unstructured document type, no template setup
  • Bulk upload — drop one file or hundreds

Bulk upload

IMGinvoice_scan.jpg
PDFpo_batch.pdf
PDFepc_report.pdf
IMGreceipt_014.png
+ drop more files, any format

02 · Reading

Reads the layout, not just the text.

ProGen keeps table and column structure intact, so line items map correctly to prices instead of turning into a wall of unstructured text.

  • Table and column structure preserved
  • Line items stay linked to the right price/quantity
  • Works on scanned and digital documents alike

Layout preserved

ItemQtyPrice

03 · Understanding

Pulls the fields that matter.

A locally-hosted LLM extracts vendor, amounts, invoice numbers, and line items, with a deterministic fallback so extraction never silently fails.

  • Vendor, amount, invoice number, and line-item extraction
  • Locally-hosted LLM, your documents don't leave your infrastructure
  • Deterministic fallback if AI inference fails

Field extraction

vendor_name
Summit Steel & Supply
invoice_number
INV-20482
total_amount
$18,420.00
Deterministic fallback active if AI inference is uncertain

04 · Generation

Generates documents, too.

Same engine, reverse direction, produce PDFs, invoices, and formatted documents straight from structured data.

  • PDF and invoice generation from structured data
  • Custom layouts and templates
  • One engine for both reading and producing documents

Document generation

Coming soon

05 · Storage & Connect

Everything searchable, everything connected.

Processed data is stored with a reference back to the original document, searchable, reviewable, and ready to push into the systems you already use.

  • Cloud storage with original-document reference
  • REST API for external systems to push and pull documents/results
  • Works toward ERP (SAP, Tally, Zoho Books), email ingestion, Slack/Teams alerts, and webhooks

Connected everywhere

Document
ERP
Email
Slack
Webhook

Stop typing in data that's already on the page.