The AWS Machine Learning Blog describes how to deploy NVIDIA’s Nemotron 3.5 Super VL, a 120B-parameter mixture-of-experts model that combines reasoning, tool use, and multimodal understanding. The post covers the model’s architecture and use cases before walking through deployment on Amazon EKS, EKS Auto Mode, and Amazon SageMaker HyperPod EKS, using NVIDIA Dynamo and vLLM for serving.
Why it matters
The guidance targets agentic AI workloads, pairing a large multimodal model with Kubernetes-based infrastructure on AWS. It outlines practical deployment paths for teams wanting to run the model within their own environments.
Who should care
ML engineers and platform teams working with Amazon EKS or SageMaker HyperPod who are evaluating multimodal, tool-using models for agentic applications.