Datalab trains specialized AI models for document intelligence, focusing on OCR, layout analysis, and PDF to markdown conversion. The company develops smaller foundation models (100-500M parameters) capable of transforming complex documents into machine-readable structured data at scale, while running on consumer-grade GPUs.

Their open-source models are state-of-the-art, easy to use, and have been adopted by hundreds of teams and researchers at leading institutions like Anthropic, Harvard, Stanford, and MIT.

Raised $3.5M seed round led by Pebblebed, with participation from Peak XV and angels including Balaji Srinivasan, Jeff Hammerbacher, and founding members of Hugging Face.