What Labelbox does
Labelbox provides data, environments, and evaluation infrastructure for AI teams, positioning itself as 'the RL data engine for AI teams.' For frontier labs it offers Horizon (reinforcement-learning environments), Terra (robotics foundation-model data), and Alignerr, a network of millions of vetted domain experts who produce preference, evaluation, and grounding signals for RLHF and alignment. For enterprises, its Recursion platform combines data integration, scenario generation, and RL training to build specialist AI agents.
Scale and backing
Labelbox started as a leading data-labeling platform and has evolved toward high-quality human data and RL infrastructure for model training and evaluation. Founded in 2018 in San Francisco by Manu Sharma, Brian Rieger, and Dan Rasmuson, the company has raised about $189M across five rounds, including a $110M Series D led by SoftBank Vision Fund, reaching roughly unicorn-level valuation. It counts a large share of leading AI labs and enterprises among its customers, with a Python SDK and APIs for programmatic data pipelines.