New serverless customization in Amazon SageMaker AI accelerates model fine-tuning

This article announces serverless customization capabilities in Amazon SageMaker AI for fine-tuning popular AI models with minimal infrastructure management.

• Supports fine-tuning for Amazon Nova, DeepSeek, GPT-OSS, Llama, and Qwen models
• Offers latest techniques: Supervised Fine-Tuning, Direct Preference Optimization, RLVR, and RLAIF
• Accelerates customization from months to days with few clicks
• Automatic compute resource provisioning based on model and data size
• UI-based customization with hyperparameter configuration and experiment tracking
• Code-based customization with sample notebooks in JupyterLab
• Deploy to Amazon Bedrock for serverless inference or SageMaker endpoints
• Built-in model evaluation and comparison against base models
• Available in US East, US West, Asia Pacific Tokyo, and Europe Ireland regions
• Pay-per-token pricing for training and inference

SageMaker AI serverless customization simplifies model fine-tuning by eliminating infrastructure management while providing flexible deployment options.