krummholzBespoke dataset projects
Your existing data.
A dataset built for the task.
Bespoke datasets from client-provided material, scoped around the records you hold, the task you want to test and the rights you can grant.
More than an off-the-shelf package.
Bespoke datasets are priced above the $14.99 catalog releases and quoted per project. The quote depends on the material, cleaning and annotation work, output format, review and delivery requirements. Applicable tax is considered for the agreed work.
Make the records useful.
A project can include schema design, cleaning and deduplication, documented transformations, labeling, train/evaluation separation, quality checks and usable CSV or JSONL exports.
Agree the permissions first.
We review the source terms, confidentiality needs and permitted use with you. Client data is not added to public datasets or used for model training without separate express permission. Retention, access and delivery are agreed for the project.
Start with a description.
Tell us what the records represent, roughly how much material you have, the task, the desired format and any timing or confidentiality constraints. Keep confidential data, personal records and credentials out of the initial enquiry. We can agree a suitable transfer process after scoping.
Bespoke dataset preparation is separate from bespoke model training. A project enquiry does not commit either party to a purchase or reserve compute.