Licensed expert data, with the paper trail.
Gridfile runs standing acquisition calls inside professional industries. What we hold was written by practitioners in the course of their work, licensed at acceptance, and delivered with a provenance record on every file.
What you receive
- A data card per call
- Domain, format, de-identification standard, review method and sampling rate, editor credentials, acceptance and return rates.
- A provenance record per file
- Agreement type and version, attestation hash, review outcomes, withdrawal status. Append-only.
- The license chain
- Contributor license to Gridfile, Gridfile license to you. Both documents, both versions, per file.
- Formats
- JSONL and Parquet, with the original file alongside. Delivered to your bucket or ours.
How we work with labs
Calls run continuously, so most requests are met from files already accepted. Where a lab needs a domain we do not yet run a call in, we write the call, recruit the editor, and open it on the grid. Volume in a new domain typically begins arriving within three weeks.
We do not produce data on commission. Everything we deliver existed before the call, which is what makes the provenance record meaningful.
Built by the team whose systems power YC-backed companies.
Talk to us
Tell us what you are training or evaluating and in which domains. We reply with what we hold and what we can open.