LABELBOX

Manu Sharma, Dan Rasmuson, and Brian Rieger were working in AI and noticed that the bottleneck was never the algorithm — it was the data. Building high-quality labeled datasets was manual, error-prone, and expensive.

They founded Labelbox in 2018 to build a collaborative platform where teams could efficiently label training data for AI models — images, text, video, geospatial data, and more. The platform also handles model-assisted labeling (using AI to speed up the labeling process), quality assurance, and dataset management.

SaaS platform with usage-based pricing. Free tier for small projects.

Enterprise plans with custom pricing based on data volume, number of labelers, and features. Revenue scales with the size and complexity of AI training data projects.

The biggest clients are enterprise companies and government agencies building AI systems.

Labelbox Annotate (data labeling tools). Labelbox Model (model-assisted labeling).

Labelbox Catalog (dataset management). Labelbox Boost (outsourced labeling services).

Support for images, video, text, geospatial, and 3D point cloud data.

Riding the AI wave. Every company building AI needs training data, and Labelbox positions itself as the standard platform for managing it.

The company grew by focusing on enterprise clients in high-value verticals: autonomous vehicles, healthcare imaging, defense, and agriculture. Partnerships with cloud providers (AWS, Google Cloud) drive distribution.

Competition from Scale AI (which is much larger and better-funded) and open-source labeling tools. The data labeling market is being disrupted by foundation models that require less labeled data (few-shot and zero-shot learning), which could reduce the total addressable market over time.

Backed by Andreessen Horowitz, SoftBank, B Capital Group, and Gradient Ventures (Google's AI fund). a16z led the $110 million Series D in 2022.