Datalab is an applied document intelligence research and development lab based in Manhattan, New York. It trains high-precision models that turn complex documents into structured, dependable data for AI systems and automated workflows. Its products include Chandra for converting PDFs, spreadsheets, slides, and other files into markdown, HTML, and JSON, with features such as word-level bounding boxes. Extraction tools identify information using defined schemas, while segmentation separates documents combined in a single file and evaluation tools assess output quality.

Organizations can compose these processors into pipelines, deploy them through Datalab’s managed cloud, within a private cloud, or in air-gapped, on-premises environments, and monitor quality through continuous evaluations. Managed batch processing supports workloads at scale, with customers paying per page processed. Datalab also develops the open-source models Marker, Surya, and Chandra, alongside a platform offering managed infrastructure and proprietary models. Its technology serves AI research, science, financial services, healthcare, and insurance, supporting use cases from training-data preparation and RAG indexing to audit and compliance.