Zoom, NVIDIA working together to deploy hybrid AI for enterprise communication

Zoom Video Communications is working with NVIDIA to integrate a hybrid generative AI architecture into its AI Companion, designed to accelerate enterprise communication workflows. The integration combines Zoom’s federated AI approach with NVIDIA’s Nemotron open-source models for task-specific, scalable AI performance. The new hybrid language model architecture intelligently switches between Zoom’s proprietary Small Language Models (SLMs), optimized for low-latency tasks, and Large Language Models (LLMs), which are fine-tuned for complex reasoning.

This enables real-time AI services such as transcription, translation, and summarisation to operate more efficiently across enterprise platforms.

Zoom’s federated model selection framework has now been enhanced with a 49-billion-parameter LLM developed using the NVIDIA Nemotron and NeMo tools, the company stated in a press release. This setup helps balance speed, cost, and output quality for AI Companion, which is increasingly used across enterprise communication systems, including Microsoft Teams, Google Workspace, and Slack.

“We’ve increased our speed and enhanced lower-cost model decision making using NVIDIA GPUs and an AI software stack, helping to optimise AI Companion’s core capabilities and enable faster go-to-market timelines,” said XD Huang, Chief Technology Officer at Zoom.

Zoom has stated that the company does not use any customer content—audio, video, chat, or files—to train its or third-party AI models. This approach is intended to meet the growing need for responsible AI across sectors with strict data privacy requirements, including finance, healthcare, and government.

“The integration of NVIDIA Nemotron into Zoom AI Companion allows enterprises like NVIDIA to experience a private, powerful, and personalised work environment for enhanced productivity,” said Kari Briski, Vice President of generative AI software at NVIDIA.