Amber

For labs

Data from real work, with the context intact.

Models learn to do a job from the record of people doing it. We source that record from companies in the domains you need, scoped to your spec.

Workflows

Tickets, email threads, docs and chat from real teams, kept in their original sequence.

Code

Repositories with PR history, reviews, CI logs and terminal sessions.

Expert files

Financial models, CAD drawings, legal drafts and other domain work product.

Recordings

Screen and voice recordings of people doing the task end to end.

Environments

Tasks, graders and reference solutions, built with domain experts.

1. Tell us the spec. Domain, data type, volume and the tasks you want to train or evaluate.
2. We source and scope. We return matching sources with size, source apps and an indicative price.
3. Review a sample. Check the data before you license it.
4. License and deliver. One contract per dataset, delivered in the format your pipeline expects.

For companies

The records you already hold have a price.

Years of tickets, documents, code and recordings are what labs want to license. We value it, remove personal data, prepare it, and run the contract. You are paid when a lab licenses it.

1. Tell us what you hold. Systems, rough size and how far back it goes.
2. Get an indicative price. We tell you what it is worth and who is likely to buy.
3. Send a sample. We return a firm offer.
4. Get paid. We handle preparation, buyer contracts and delivery.

Partners

Know a company sitting on valuable data?

Introduce us. We run the deal from the first call to delivery.

How we handle data

Licensed, and cleaned before delivery.

• Every dataset is licensed from its owner.
• Personal data is removed before delivery.

Contact

Tell us what you need.

We read every message and reply by email.