Retab Secures $3.5M Pre-Seed to Power Reliable Document AI
San Francisco-based AI startup Retab has successfully closed a pre-seed funding round, securing $3.5 million to advance its mission of making document AI truly production-ready for developers. The investment round was led by prominent early-stage funds VentureFriends, Kima Ventures, and K5 Global, with additional backing from notable angel investors including former Google CEO Eric Schmidt (via StemAI), Datadog CEO Olivier Pomel, and Dataiku CEO Florian Douetteau. This significant capital injection coincides with the official launch of Retab’s innovative platform, designed to tackle the pervasive challenges plaguing document automation.
At its core, Retab addresses what its founders call the “broken state of document AI,” where impressive demonstrations often fail to deliver reliable performance in real-world production environments. Many developers face the arduous task of piecing together fragile pipelines just to extract basic information from documents like PDFs. Retab aims to be the operating system for reliably extracting structured data, acting as an essential intelligence layer that sits between popular large language models (LLMs) from providers like OpenAI, Google, and Anthropic, and the vast amounts of unstructured data held within enterprises. Instead of being another LLM, Retab focuses on making existing powerful models usable for critical workflows by transforming messy documents, including handwritten scans, into clean, structured data.
The platform offers a developer-first toolkit that automates the complex lifecycle of document processing. Developers simply define the schema of the data they need, and Retab handles the intricate details, from dataset labeling and evaluations to automated prompt engineering and optimal model selection. This approach is bolstered by a system of intelligent checks and balances, including self-optimizing schemas that refine instructions for maximum accuracy, intelligent model routing that assigns tasks to the best-performing models based on cost, speed, or precision, and guided reasoning with k-LLM consensus to enhance reliability by coordinating outputs across multiple models.
Founded by engineers Louis de Benoist (CEO), Sacha Ichbiah, and Victor Plaisance, who hail from prestigious institutions like Cambridge and École Polytechnique, Retab’s origins stem from their firsthand experience building internal automation tools for document-heavy workflows, particularly in logistics. They realized the true value lay not just in the output, but in the robust orchestration layer they had built to make AI models function effectively. This tooling became the foundation for Retab, which is already being utilized by dozens of companies across diverse sectors such as logistics, finance, and healthcare to extract structured data from complex, real-world inputs.
The strong syndicate of investors underscores the market’s need for Retab’s solution. VentureFriends, an Athens-based early-stage VC, is known for its founder-first approach and investments across Europe and MENA. Kima Ventures, a highly active Paris-based firm, invests globally in early-stage startups, often leading pre-seed and seed rounds. K5 Global, a San Francisco-based VC and incubation studio, focuses on Enterprise SaaS, Vertical SaaS, and FinTech. The participation of tech luminaries like Eric Schmidt, Olivier Pomel, and Florian Douetteau further validates Retab’s vision, with Douetteau noting that the “AI-fication of the economy depends on the capability to convert operations based on millions of documents into verified, structured data that autonomous systems can utilize,” a challenge Retab is uniquely positioned to solve.
Looking ahead, Retab plans to expand its capabilities beyond documents to include data extraction from websites and integrate with popular automation platforms like n8n, Zapier, and Dify. This strategic expansion positions Retab as intelligent middleware, connecting AI agents to the vast, often impenetrable, world of unstructured data that powers global operations, ultimately striving to make advanced AI truly dependable for enterprise-scale applications.