NuExtract: The Sovereign AI for Document Extraction
Dec 4, 2026 | 2:00 PM - 2:10 PMStartup Stage
Dec 4, 2026 | 2:00 PM - 2:10 PM
Startup Stage
Description
Accurately extracting information from documents is key to automating business processes and providing reliable context for AI agents.
Traditional OCR solutions are increasingly being replaced by multimodal LLMs. However, these models remain error-prone, are expensive to operate, and usually require sensitive documents to be shared with third-party services.
I will present NuExtract, a compact VLM designed specifically for document extraction. It outperforms general-purpose models while remaining small enough to deploy on-premises, offering a fully sovereign solution.
NuExtract is deployed in document-intensive enterprises, including AXA, for both structured extraction (documents → JSON) and content extraction (document → Markdown).




