companies

Labelbox

Data labeling and training-data platform that has shifted from computer-vision annotation to reinforcement-learning data for frontier models.

Labelbox was founded in 2018 by Manu Sharma, Brian Rieger, and Daniel Rasmuson to solve a problem every early computer-vision team hit: labeling images and video consistently enough that a model could actually learn from them. The platform grew into a broader annotation tool covering text, audio, and documents, with consensus scoring, review workflows, and model-assisted pre-labeling to cut down the manual grind, and it picked up customers ranging from Walmart to Genentech along the way. More recently the company has repositioned around reinforcement learning data for large language models, describing itself as running "data factories" that supply human feedback and preference data to labs building frontier models. Backed by Andreessen Horowitz, SoftBank's Vision Fund, and Databricks Ventures among others, Labelbox closed a $110 million round in early 2022 that pushed its valuation past $1 billion, and it continues to compete with Scale AI and Surge AI for the large data-labeling contracts that now underpin model post-training as much as they once underpinned vision models.

Founded

2018

Headquarters

San Francisco, United States

Sector

data