Turn an unreadable archive into structured, cited evidence.
Existing buildings don't have a data problem — they have a legibility problem. The evidence usually exists, spread across handover packs, scanned certificates and drawings that predate any requirement to keep them. Veridatum reads all of it and returns structured facts, each one traceable to the page it came from.
What it does
Bulk archive ingestion
Point Veridatum at the whole estate — a SharePoint export, a network drive, scanned boxes. Every file is fingerprinted, de-duplicated and registered. Nothing is silently skipped: formats we can't parse still appear in the register as a coverage gap.
Classification by content, never filename
Each document is typed against a controlled vocabulary — fire strategy, FRAEW, commissioning certificate, as-built drawing — judged on what it actually contains. The most safety-critical certificate in an archive is usually the one nobody named properly.
Extraction with a citation on every fact
Safety-critical attributes are pulled against a canonical schema, and each value stores the verbatim quote and page that evidences it. A fact that can't be cited is discarded rather than guessed.
Legacy drawing recovery
Blueline prints, fold creases, imperial dimensions in pencil, stamp overprints. Sheets are triaged cheaply, then information-dense regions are re-read at high resolution, with values corroborated across sheets before they're treated as confident.
What you end up with
- A structured record built from documents nobody has been able to search
- Every value traceable to a source document and page
- Scanned and handwritten material included, not skipped
The rest of the platform
These areas are designed to be used together — the register is only as good as the extraction behind it.
See it on a real building
The demo runs on an anonymised 15-storey tower, with the findings the pipeline actually produced.