Active Learning: Improve your models with production data

You can now configure and manage continuous model improvement from a revamped Active Learning inside your project. It surfaces incoming images, collection batches, reviewers, conditions, and filters, so you can set collection limits, define what gets captured, and assign reviews in one place.

Active Learning can be enabled directly from a Workflow, configured through the REST API, and managed through the MCP Server, so an agent like Claude can set it up for you. You can also start an empty project with a foundational or open source model and begin collecting and reviewing annotated data from day one, then bring in your fine-tuned models as they are ready.

Learn more in the documention