Retab raises $3.5M to launch developer-first platform for revolutionizing document AI

Founded by engineers from logistics automation, Retab offers a developer-first platform that industrializes document processing with production-grade reliability, managing the full data extraction lifecycle—from schema definition to model selection—for vertical AI startups and enterprise teams

San Francisco-based startup Retab has announced $3.5 million in pre-seed funding and the launch of its document intelligence platform aimed at solving one of the most overlooked bottlenecks in AI—understanding unstructured documents. The round saw participation from prominent investors including VentureFriends, Kima Ventures, K5 Global, and notable figures like Eric Schmidt (via StemAI), Olivier Pomel (CEO, Datadog), and Florian Douetteau (CEO, Dataiku).

Founded by engineers who previously built automation tools for document-heavy logistics workflows, Retab aims to streamline and industrialize document processing for vertical AI startups and enterprise teams. Unlike many AI tools that falter in real-world deployment, Retab promises production-grade reliability through its intelligent orchestration layer.

Retab is not a large language model (LLM) itself, but a developer-first SDK and platform that handles the entire data extraction lifecycle. Developers define what structured data they need, and Retab manages everything else—from dataset labeling and prompt optimization to model selection and performance benchmarking.

"People build AI demos that fall apart in production," said Louis de Benoist, Retab’s co-founder and CEO. “We created Retab to be the platform we always needed—developer-friendly and production-ready.”

Automation, accuracy, and AI readiness

Key features include Self-Optimizing Schemas, which test and refine extraction logic before going live; Intelligent Model Routing, selecting the most efficient model based on cost, speed, or accuracy; and a Guided Reasoning system that forces LLMs to think step-by-step with multi-model consensus for reliable results.

Already in use across industries like logistics, finance, and healthcare, Retab is enabling automation of complex tasks such as claims processing, quarterly report analysis, and identity verification—with minimal setup and up to 100x cost efficiency.

Dataiku CEO Florian Douetteau emphasized Retab’s significance, stating, “The AI-fication of the economy depends on converting document-heavy operations into structured, reliable data. Retab is uniquely positioned to power that transformation.”

Looking ahead, Retab plans to expand its extraction capabilities to websites and integrate with automation platforms like n8n, Zapier, and Dify. With just 10 employees and growing developer traction, Retab is emerging as a vital middleware layer between unstructured data and the AI agents that rely on it.