Amazon Web Services has announced new tools for enterprise customers to create their own frontier AI models, following closely on the launch of its Nova Forge service for training custom Nova models.
The cloud provider unveiled new capabilities in Amazon Bedrock and Amazon SageMaker AI at its AWS re:Invent conference. These features are designed to make building and fine-tuning custom large language models easier for developers.
A key introduction is serverless model customization in SageMaker. This allows developers to begin building a model without managing compute resources or infrastructure. Developers can access these capabilities through a self-guided point-and-click path or an agent-led experience using natural language prompts, with the agent-led feature launching in preview.
As explained by Ankur Mehrotra, general manager of AI platforms at AWS, a healthcare customer could use this to make a model understand specific medical terminology better. By providing labeled data and selecting a technique, SageMaker would then fine-tune the model automatically. This capability works for customizing Amazon’s own Nova models and certain open source models like DeepSeek and Meta’s Llama.
AWS is also launching Reinforcement Fine-Tuning in Bedrock. This allows developers to choose either a reward function or a pre-set workflow, and Bedrock will run the entire model customization process from start to finish.
Focusing on frontier models and customization appears to be a strategic priority for AWS at this year’s conference. This follows the announcement of Nova Forge, a service where AWS builds custom Nova models for enterprise customers for an annual fee.
Mehrotra noted that customers are seeking ways to differentiate their solutions when competitors have access to the same base models. The key to solving that, according to AWS, is the ability to create customized models optimized for a specific brand, data, and use case.
While AWS has yet to gain a substantial user base for its AI models compared to providers like Anthropic, OpenAI, and Gemini, the company believes its enhanced customization and fine-tuning tools could provide a competitive advantage moving forward.