About

A marketplace built around evidence, not claims.

ModelData Exchange connects AI companies that need software engineering data with the businesses and developers who own it and can license it legitimately.

The model

Two sides, one review in the middle

Our team sits between data owners and AI companies. Nothing is listed or delivered until rights and quality have each been reviewed and recorded.

Data ownersSuppliersCompanies, agencies, and developers who own engineering records and choose how they are licensed.
ModelData ExchangeManaged review
  • Rights review
  • Technical QA
  • Preparation
  • Contracts & delivery
AI companiesBuyersTeams training and evaluating coding models and agents who need data they can explain.
A written rights record and quality report travel with every dataset version.

Why we exist

AI coding systems learn from how software is built: problems reported, changes proposed, reviews given, tests passed. Much of the most useful record of that work sits inside companies, where it is rarely available — and when it is, its provenance is often unclear.

Public code is not a free-for-all. Licences differ, contributors have agreements, and employers and clients have rights of their own. Teams that train models on data they cannot explain take on risk they may not see.

We think there is a better path: data sourced from owners who agree to license it, reviewed for rights and quality before anyone pays for it, and licensed on terms both sides understand.

How we operate

We run a managed marketplace. Our team qualifies buyers and suppliers, reviews rights and provenance, checks quality, and coordinates proposals, contracts, and delivery. Software supports that work; it doesn’t replace judgement.

What we don’t do

  • Accept source code or repositories through public forms
  • Assume public or open-source code can be relicensed for any use
  • Offer self-serve checkout or automated pricing
  • Claim certifications, customers, or results we can’t show

See the full process →

Principles

How we make decisions

  1. Evidence over assertionEvery claim about a dataset — who owns it, what it contains, how it was checked — is backed by a record you can read.
  2. Owners decide how their data is usedSuppliers choose the uses they license, the exclusivity they offer, and can withdraw from new licensing at any time.
  3. Honest descriptions, including limitationsDataset cards say what the data is, what we checked, and what we did not. Unverified numbers are not shown.
  4. No promises we can’t keepWe do not promise earnings, model performance, or legal outcomes. A review reduces risk; it is not a guarantee.
  5. Security proportionate to trustWe handle other people’s engineering records carefully, and describe our controls as they are today.

Talk to us.

Questions about ModelData Exchange, our process, or whether your data is a fit? We’ll answer plainly.