About Labelbox
Labelbox is a platform for building AI training data and environments. Its newer work centers on agents that handle recurring engineering and research jobs.
What it does
Labelbox provides the data and training environments that AI teams use to build models. Its Recursion product runs agents on code, infrastructure and experiments, and those agents can start from events such as a failed build, an alert or a schedule. The agents work in the background and send their results for human review before anything is merged.
Key features
- Training environments for model builders
- Background agents for engineering and research work
- Jobs triggered by builds, alerts or schedules
- Fixes sent to pull request review and team chat
Who it's for
AI labs building frontier models, and engineering or research teams with a steady backlog of technical tasks. Teams that already work with GitHub, Jira or Slack may find the agent tools easier to fit into their routine.
Pricing and good to know
Pricing is set by the vendor, and enterprise access is offered for larger teams. The agents connect to outside tools such as code hosts and chat apps, so check the permissions they receive before setup. Results depend on how well the team describes the jobs it wants handled.
Highlights
- Build training environments for AI models
- Run agents that work on code and infrastructure
- Start agent jobs from builds, alerts or schedules
- Send agent fixes to pull request review
