Deasy Labs

Deasy Labs is a context engine that turns unstructured data — including SharePoint sites, emails, PDFs, and cloud storage — into curated datasets for AI projects. It addresses the data preparation step that precedes retrieval and agent workflows: mapping what exists, filtering out sensitive, irrelevant, or low-quality files, and enriching content with metadata so AI systems can retrieve the right information.

The platform connects to sources like SharePoint, S3, and existing vector databases, then ingests, OCRs, parses, chunks, and normalizes content. It tags files using LLMs and ML models, supports custom or auto-generated taxonomies, and applies relevance, quality, and sensitivity scoring. Users can slice data by topic, time, quality, or sensitivity and publish datasets to RAG pipelines and retrieval systems, or write metadata back to source systems. Deasy continuously monitors sources and refreshes datasets as new content arrives. It is available as APIs or a no-code platform, with integrations including Google Cloud Vertex and Gemini, LlamaIndex, and Qdrant, and supports deployment in your own environment.

Deasy is aimed at data and AI teams building RAG, search, and agent systems, as well as engineers and data scientists managing unstructured data at scale. Deasy Labs was acquired by Collibra in 2025.