Amazon SageMaker AI now supports serverless model customization for NVIDIA Nemotron 3.5 Lightning
Amazon SageMaker AI enables serverless customization of the NVIDIA Nemotron 3.5 Lightning model using SFT, DPO, and RFT, allowing domain-specific adaptation without infrastructure management.
Amazon SageMaker AI now supports serverless model customization for the NVIDIA Nemotron 3.5 Lightning model, expanding its capabilities beyond standard deployment. The model employs a hybrid Mixture-of-Experts architecture with 3 billion active parameters and 30 billion total parameters, offering a balance between efficiency and performance. Users can adapt the model to proprietary datasets using supervised fine-tuning, Direct Preference Optimization, or reinforcement fine-tuning to align outputs with specific requirements. This approach reduces operational overhead by eliminating the need for manual infrastructure provisioning or cluster management during training.
Serverless customization allows organizations to tailor the model to their workflows while paying only for the compute resources consumed during the process. The feature supports labeled data for accuracy improvements, preference data for tone alignment, and reward signals for performance enhancement on new tasks. SageMaker AI handles all orchestration, enabling teams to focus on data preparation and evaluation rather than technical setup. This reduces both cost and latency compared to traditional fine-tuning methods.
The serverless model customization feature is available in multiple AWS regions, including US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo), and Europe (Ireland). Users can initiate customization jobs through the Amazon SageMaker Studio Models page or programmatically via the SageMaker Python SDK. The process streamlines access to advanced model adaptation without requiring specialized infrastructure expertise.
To begin using serverless model customization, users can navigate to the Models page in Amazon SageMaker Studio or leverage the SageMaker Python SDK for automated workflows. Detailed guidance is provided in the Amazon SageMaker AI model customization documentation, which outlines supported methods and best practices for implementation.