Datalab

Datalab is a research lab that trains document intelligence models and sells a hosted API for converting and extracting data from documents. The API covers Convert (PDFs, spreadsheets and slides to markdown, HTML and JSON), Extract (user-defined fields returned as JSON with citations and a confidence score per field), Segment (splitting documents combined into one file) and Eval (comparing processing modes on a customer's own documents). Processors can be chained into versioned, immutable pipelines that are called by ID. The company also publishes open-source models, including Marker, Surya, Chandra and Indra.

Customers named on the website include AI research labs, financial services, healthcare and insurance firms, and enterprises. Revenue comes from pay-as-you-go API usage charged per processor, a Team plan at $400 per month, and Enterprise contracts with custom terms, including deployment in a customer's own VPC or fully air-gapped on-premise. The company states that its open-source models have been adopted by teams including Anthropic, ElevenLabs and Legora.

Datalab operates from an office in Manhattan, New York. It offers SOC 2 Type II compliance, BAA and DPA agreements, and EU data residency. Its current open roles are all based in the New York office: enterprise solutions engineer, founding marketer, research engineer and senior software engineer.

Product: Document parsing and extraction API (Convert, Extract, Segment, Eval) with versioned pipelines, plus open-source OCR models Marker, Surya, Chandra and Indra.