Real people. Real-world data.

Human DataBuilt For Better AI

Contribute real-world data. Train better AI.Earn for your contribution.

Built on human experience

Ways to contribute.

Capture
Photos and video of real places, objects and activity.
Upload
Documents and files you already have the right to share.
Label
Describe what is in the data, precisely.
Evaluate
Judge which model output is better, and say why.
Verify
Confirm that existing labels are correct.

The data behind better AI

Everyday human activity becomes training data.

Models learn from what they are shown. Kernel collects what the internet doesn’t have: ordinary people doing ordinary things, recorded on purpose and with consent.

Egocentric video
First-person recordings of everyday tasks: cooking, assembling, repairing.
Speech
Read and spontaneous speech across accents, ages and rooms.
Images
Objects and scenes as they actually appear in people's homes.
Documents
Receipts, forms and letters in the formats the world really uses.
Computer interactions
Screen recordings of people completing real tasks in real software.
Spatial data
Walkthroughs and multi-angle captures of rooms and objects.
Human preferences
Comparisons and ratings that tell a model what good looks like.
Need something that isn’t here?
Request custom data

Contribute → Verify → Earn

From one contribution to a better model.

Every sample in a Kernel dataset can be traced to a consented contribution, the checks it passed, and the person who reviewed it.

Contribute
Pick a mission. Capture, upload, label or evaluate from your phone or laptop.
Kernel verifies
Automated checks, then human review. Every file is hashed, every decision recorded.
Datasets are created
Accepted work is assembled into versioned datasets with full provenance.
AI gets better
Labs train and evaluate on data that reflects how people really live and work.
You earn
The reward is shown before you start and paid when your work is accepted.