Docusign Accelerates Agreement Management With NVIDIA’s Inference Platform

Docusign specializes in helping organizations create, commit to, and manage agreements. To enhance customer experiences, Docusign turned to AI to deliver new digital agreement capabilities. However, they encountered two common challenges: optimizing throughput and managing infrastructure expenses. Their shift to cloud-based infrastructure aligned with their transition to NVIDIA products, facilitated by their relationship with Microsoft Azure.

Docusign uses NVIDIA Triton Inference Server to accelerate AI model deployment and reduce overall ownership costs. Since deploying the unified AI inference server in production over a year and a half ago, Docusign has expanded its usage across various AI scenarios designed for specific agreement needs and pain points. For example, Docusign helps organizations analyze and extract more meaning from their agreements by unlocking and surfacing data and metadata stuck in agreements like identities of parties, terms and conditions, start and end dates, and more—without tedious, manual work.

Initially deployed on on-premises CPU servers, Docusign has since migrated their AI models to NVIDIA GPUs running on Microsoft Azure Cloud. This transition, coupled with serving the models using NVIDIA Triton Inference Server, has resulted in a remarkable increase in throughput. The swift adoption of AI services within the digital agreement landscape has spurred Docusign’s AI product teams to delve deeper into inference optimizations. Recently, the company embarked on compiling their models for inference using NVIDIA® TensorRT™ to further reduce latency and is transitioning to more powerful NVIDIA A100 Tensor Core GPUs, available on Azure Cloud.