Private document intelligence
The documents you can’t put in the cloud, finally searchable.
Names, addresses and account numbers are replaced with placeholders before your archive is ever indexed — so the searchable copy holds no identifiers to leak. You read answers with the real names restored from an encrypted map only you hold. Everything built on top is working on an archive that does not know whose it is.
Contracts, client files, correspondence, filings, scans and recordings, searchable in plain language — with every answer linked to the passage it came from.
Which version of the Aldridge engagement letter is current, and what changed in the liability cap?
The 12 March 2019 revision supersedes the 2004 agreement. The liability cap changed from a fixed sum to the fees paid in the preceding twelve months.
The problem
The answer is in the pile. You just can't find it.
Firms don’t lose documents — they lose the ability to search them. An indemnity clause agreed in
2009 sits in a superseded contract, a settlement position sits in an archived mailbox, a signed
variation sits in a scan named IMG_4471.jpg. When it suddenly matters, none of it is
findable — and the obvious fix, handing it all to a cloud assistant, is the one thing a firm
holding client confidences cannot do.
It reads the formats you actually hold
Text documents, spreadsheets, email, web pages, ordinary images, scanned PDFs, audio and sampled video frames are implemented. Handwritten prose gets a second, context-aware reading that is searchable but never quoted as the page’s words — measured at 95% of words on one transcribed page, against 59% without it. Handwritten forms and photographs taken in poor conditions remain under release validation.
You ask in plain language
No folder structure to maintain, no tagging discipline to keep up. Ask the question you'd ask a person who had read every page.
Answers show their source
Responses show the exact indexed passage and file identity they used. That proves where text came from, not that every generated claim is correct; the original remains available for checking. Physical page numbers are shown only where the extractor preserves them.
What it is
Four jobs, normally four vendors, normally four uploads.
Searching a confidential archive is not one problem. It is four, and each is usually solved by a different service that wants a copy of your documents. Veldrun does all four on one machine.
Read it
Scanned agreements, photographed letters, spreadsheets, email, recordings and video — read locally, so a photograph of a signed variation becomes searchable text without leaving the machine.
Take the identities out
Names, addresses and account numbers become placeholders before anything is indexed. The real values go into an encrypted map only you hold, and a document that cannot be redacted confidently is set aside rather than indexed.
Answer the question
Ask in plain language and get the exact indexed passage the answer came from, so you can check it against the original.
Weigh the sources
A signed order and an email arguing about that order look equally relevant to a similarity score. Veldrun is built to weigh which source deserves to be believed.
Availability
Veldrun is not for sale yet.
The Windows application is built and running every day on a real archive of legal and medical records. It is in limited pre-release while the last release gates are proven, and it is not on sale — nothing on this page is an offer.
No prices, plans or limits are published yet, and none can be reserved or purchased today. When commercial terms are settled they will appear here first, alongside the seller's identity and service address.
Early access
Veldrun isn't on sale yet.
Leave your email and we'll write when a public test or purchase is genuinely ready. No newsletter sequence; the address is processed only for the waitlist and required service delivery.
One email. Unsubscribe in a single click. Veldrun is a pre-release project; see the privacy policy for what is stored and how to request deletion.