Datalab is an applied research lab in New York. We train our own models to turn documents into data you can verify: scans, dense tables, handwriting, the pages other tools quietly get wrong.
Because we own the models, a customer’s hard problem becomes a roadmap conversation, not a feature request. Our open source work (Marker, Surya, Chandra) has more than 70,000 GitHub stars. Customers include Anthropic, ElevenLabs, Gamma, Reflection AI, and Legora, along with insurers, financial firms, and healthcare companies. We deploy anywhere from our API to fully air-gapped on-prem.