Unstructured

Market position

Described in press coverage as a pioneer in unstructured-data pre-processing for LLMs, with an analyst noting that comparable data-preparation tooling exists but remains lightly adopted because much enterprise work is still manual. The company has been recognized by CB Insights (AI 100), Forbes (Top 50 AI Companies), Fast Company (#24 Most Innovative) and Gartner (Cool Vendor 2024).

Breadth of file-type coverage (64+ formats) and connector coverage (30+ sources and destinations) combined with maintained, production-grade pipelines, an extensible Python plugin architecture, and built-in security, compliance and role-based access control. Customers cited on the company site highlight accuracy on complex documents, reduction of internal data-plumbing engineering, and availability across multi-cloud and on-premises environments.

Technology

Modular partitioning functions and connectors that ingest images and text documents (PDF, HTML, Word, email, XML, JSON and more) and emit structured elements; a Python-based plugin architecture that customers can extend; chunking, enrichment and embedding stages; integrations with model providers such as OpenAI and Anthropic; Docker images built on wolfi-base for x86_64 and Apple silicon; and an MCP server exposing document processing to agent frameworks. The enterprise platform is described as continuously extracting data from existing databases and transforming it into LLM-ready formats in near real time before loading into vector stores.

Go-to-market

Bottom-up developer adoption of the open-source library and documentation, converting into paid API and platform usage, combined with direct enterprise sales (book-a-demo, contact-sales motions) and partner/channel distribution through data infrastructure vendors such as Teradata, IBM, SAP and Databricks. Distribution also occurs through MCP clients including Claude Code, Cursor and Codex.

Enterprise data, platform and AI engineering teams building GenAI, RAG and analytics applications, particularly in complex and regulated data environments; also independent developers and agent builders using the open-source library and MCP server. Named or quoted users include Teradata, IBM, SAP, Uniphore and Intel; the company claims usage across 87% of the Fortune 1000, and 2024 reporting cited more than a third of the Fortune 500 among library users.

Geography

Headquarters reported as San Francisco, California; the founder is described as a Sacramento native and the company has been profiled by the Sacramento startup community. Deployment is described as spanning multi-cloud and on-premises environments, with a customer (SAP) using it across a global GenAI stack.

History

Technology was developed in collaboration with the open-source community, commercial enterprises and U.S. government defense and intelligence organizations, supported by Air Force and Space Force SBIR Phase I/II contracts and additional backing from U.S. Special Operations Command. A $25 million raise was announced in July 2023, followed by a $40 million Series B led by Menlo Ventures in March 2024, bringing disclosed funding to more than $65 million. The commercial SaaS API launched in January 2024 and the enterprise platform in February 2024. An aggregator source reports a further $40 million Series C at a $200 million valuation dated October 2025. Product scope has since expanded to a no-code UI, hosted Pipelines, Foundation knowledge layers and a Transform MCP server for agents.

Risks & controversies

Sources disagree on basic company facts: SiliconANGLE reports a 2022 founding while the SalesTools aggregator reports 2020, and the aggregator's Series C report is a low-detail secondary source not corroborated elsewhere in the material. Traction claims such as 87% Fortune 1000 usage originate from the company's own website and are unverified. An outside analyst quoted in coverage notes that data-preparation tooling faces competition and limited enterprise adoption, with much work still done manually.