Welcome to Datalab

Datalab builds state-of-the-art document intelligence models to convert complex PDFs and other unstructured formats into structured, machine-readable outputs — fast, accurately, and at scale. At our core, we provide:

• Marker: Our open-source document processing system that converts documents to structured formats with high speed and accuracy.
• Surya: Our comprehensive document OCR toolkit designed for processing various document types with capabilities that include text detection, text recognition, layout analysis, reading order determination, table recognition, and LaTeX.

​

Getting started

Whether you want to securely host us in your own environments, or use our hosted API, we make it easy to get started.

​

What can I use Datalab for?

Datalab is built for anyone working with messy, high-stakes, or high-volume documents. Our users span industries, teams, and use cases. Some examples include:

• AI/ML teams building agents or structured data pipelines: Feed RAG systems, knowledge graphs, or automation workflows with clean, structured outputs. Ideal for converting unstructured PDFs into JSON, Markdown, or HTML for downstream use.
• Enterprises with compliance-heavy document processing needs: Automate high-volume document review and extraction with auditability, bounding boxes, and deterministic parsing.
• Product or Engineering teams in EdTechs, legaltechs, fintechs, and AI research labs: Turn scanned textbooks, legal filings, financial statements, tax forms, research papers, into product-ready content at scale.

​

Support