From the model's first prediction to the expert's final correction, Kili keeps every version of every judgment and the name behind it. Whether you're building training data or evaluating outputs, the record of how your data was made is there before anyone asks for it.
.svg)
.svg)
.svg)
Submitting work in Kili turns it into a version. When a labeler revises after submitting, or an admin corrects a result without adding a review step, the change becomes a new version under their name, with the date and time. The original stays on the asset, still credited to whoever made it.
Open any asset in Explore and see what shaped it: the model prediction it started from, then each review and correction, every one named and timestamped. Versions that were superseded stay visible.
Coming soon: compare two versions side by side or as an overlay, with each addition, edit and deletion listed.
Coming soon. The audit trail records the actions that shape a project, from role changes and ontology edits to workflow changes, label revisions and data exports. Each one is dated, attributed, and linked to its asset where there is one. Filter it to look into an incident, or export it as JSON for your auditors.
Speak to our team for a free 14-day evaluation.
.webp)
When model outputs are imported into Kili, each one becomes the first version on its asset. Your experts accept it, correct it or send it back, and every step adds a new version with their name and the time, while the model's original stays on record. When someone asks whether a person actually checked the outputs, you can show who did and what they changed.
Explore AI Evaluation
Post-training data often starts as a model's draft that specialists rewrite. Kili keeps all of it: the draft as it arrived, the specialist's revision and the reviewer's sign-off, each a separate version with a name on it. When criteria change and an admin corrects a response, that becomes one more version, and the earlier ones are still there.
Explore Post-Training
In a regulated file, a label is a decision and needs an owner. Each annotation is saved under the person who made it, a review lands as its own version instead of an overwrite, and Enforce Step Separation keeps labeler and reviewer apart by default. Where several annotators label the same asset, you can see how far they agreed next to the work.
Explore Data Labeling
Kili Technology is a complete data suite that supports all data types and handles specialized formats for domain-specific requirements.
Customer stories from the programs that put this to work — with the numbers behind them.
.png)
.png)
.png)
Stay up to date with fresh content from our team — tutorials, use cases, and ideas to help you train AI/ML models better.
Got questions about Kili Technology? Check out our FAQ. If you can't find it in this list, drop a question for our team.
%3Aformat(webp).webp)
%3Aformat(webp).webp)
.png)
.png)
