Axios Pro Exclusive Content

Exclusive: Sapien raises $5M to ease AI's data friction point

headshot
Apr 10, 2024
Illustration of a robotic hand holding a glowing dollar sign.

Illustration: Allie Carl/Axios

Sapien, a company that can help retailers contend with unstructured data, raised a $5 million seed, co-founders Henry Chen and Kelly Ryan tell Axios exclusively.

Why it matters: Clean and properly labeled data helps companies train AI models more precisely.

Zoom in: Primitive Ventures, Animoca, Ravikant Capital and Yield Guild Games participated in the round.

The big picture: Many companies rely on humans to parse through and label large swathes of unstructured data — which, "at its core is a very boring task," Chen says.

How it works: Sapien gamifies the data labeling process, letting workers earn points for blockchain-based rewards like tokens and NFTs in exchange for correctly and efficiently labeling images, texts and other data.

  • "Fine-tuning" — the process of adapting AI models to new data and tasks — is integral to building effective AI models for chatbots, image recognition, or search, Ryan says.

What they're saying: "Having the infrastructure in place to have a human go through, produce the labels and know that they're correct enough to actually train your model on is a pretty hard problem," Ryan says.

What's next: Sapien wants to expand to more diverse datasets and new industries, like health care, supply chain and education tech, Chen says.

Go deeper