NVIDIA-Certified Associate Generative AI Multimodal

A data scientist begins prototyping a generative AI project on a local desktop workstation but anticipates that model training will soon require more compute than the local GPU can provide. Which characteristic of NVIDIA AI Workbench best supports this workflow?

A team wants to rapidly prototype a RAG-powered chatbot where the embedding step runs locally, but inference can be performed either locally or in the cloud. Which AI Workbench chatbot approach directly matches this requirement?