Retab, a developer-focused AI startup, has emerged from stealth with €3 million in pre-seed funding and the official launch of its document automation platform.
The round was backed by VentureFriends, Kima Ventures, and K5 Global, along with angel investors including Eric Schmidt (via StemAI), Datadog CEO Olivier Pomel, and Dataiku CEO Florian Douetteau.
“The OS for structured data”: AI pipelines made production-ready
Founded by engineers Louis de Benoist, Sacha Ichbiah, and Victor Plaisance, Retab helps developers turn unstructured documents — from scanned PDFs to handwritten forms — into production-grade structured data. The platform automates extraction pipelines, model routing, evaluation, and dataset labeling.
Retab acts as an AI orchestration layer, selecting the best model for each task from providers like OpenAI, Anthropic, and Google, and optimizing output with tools like k-LLM consensus and guided reasoning.
Advertisement
Serving vertical AI teams with real-world production needs
“Most LLM demos break in production,” said CEO Louis de Benoist. “Retab wraps top models in logic, error handling, and structure — so developers can actually ship real apps, not just prototypes.” The system has already been used in sectors like logistics, finance, and healthcare to automate document-heavy workflows such as claims processing, report analysis, and identity verification.
Not just an AI layer — a document infrastructure platform
Unlike lightweight no-code tools, Retab is built as a middleware layer for developers. It supports SDK integration, self-optimizing schemas, and performance benchmarking across speed, accuracy, and cost. Future releases will expand into web data extraction and add integrations with tools like Zapier, n8n, and Dify.
From internal automation tools to commercial infrastructure
The founding team previously built internal tools for automating logistics paperwork, and the frustrations they encountered led to Retab’s creation. With a team of 10 and an expanding community, the startup is positioning itself as a core component of the AI infrastructure stack — especially for vertical AI startups scaling unstructured data use cases.
