Docusign Deploys NVIDIA Nemotron Parse to Extract Complex Contract Tables at Scale

Docusign has integrated NVIDIA's Nemotron Parse, a vision-language model built for document understanding, into its production document processing pipeline to improve contract table extraction. The company, which handles millions of daily transactions across nearly 1.9 million customers in over 180 countries, has long struggled with accurately pulling structured data from complex contract tables containing merged cells, multi-page layouts, and nested formatting. General-purpose AI models failed to reliably handle these structures, often requiring costly manual corrections that slowed enterprise workflows. The solution was unveiled at the AI Engineer World's Fair in San Francisco, where Docusign and NVIDIA jointly demonstrated the system's architecture. The pipeline uses vLLM to serve the model while keeping all sensitive agreement data within Docusign's secure, on-premises environment to meet strict data governance requirements.