AWS simplifies model customisation
AWS used its re: Invent 2025 conference to detail its work focused on simplifying model customisation to help developers build faster, more efficient AI agents.
The company says that, now, Amazon Bedrock and Amazon SageMaker AI put advanced model customisation into the hands of any developer. Also, reinforcement fine-tuning in Amazon Bedrock makes it easier to tailor models to specific use cases and improve accuracy.
Amazon SageMaker AI is said to cuts advanced model customisation workflows from months to days, accelerating AI development and bringing new solutions to market faster.
Although building AI applications has become easier, AWS says that running them at scale remains expensive and resource-intensive and the challenge is particularly acute for AI agents, which can have higher inference demands as they reason through problems, use a variety of tools and coordinate across multiple systems.
Agentic challenges
Why is agentic development such a challenge?
The engineering team at AWS say that a significant amount of an agent’s time is spent doing routine tasks, like checking calendars and searching documents, that don’t require advanced intelligence… all of which results in unnecessary costs and wasted resources.
The solution, it appears, lies in customisation i.e tailoring smaller, specialised models to handle the work agents do most often to deliver faster, more accurate responses at lower costs.
But, advanced customisation techniques like reinforcement learning required deep machine learning expertise, extensive infrastructure and a lot of development time.
“We announced new Amazon Bedrock and Amazon SageMaker AI capabilities that make advanced model customisation accessible to developers at any organisation. Reinforcement Fine Tuning (RFT) in Amazon Bedrock and serverless model customisation in Amazon SageMaker AI with reinforcement learning simplify the process of creating efficient AI that’s fast, cost-effective and more accurate compared to base models. By making these techniques more accessible for our customers’ developers, we’re making it easier for organisations of all sizes to build custom agents for any business need,” notes AWS, in a product statement.
Difficult customisation techniques present a roadblock for building custom, efficient models. Reinforcement learning, for example, trains a model using feedback from either humans or another model.
“Good behaviour gets reinforced, while bad behaviour gets corrected. It’s particularly good for reasoning and complex workflows because it rewards good processes, not just good answers. However, reinforcement learning requires a complex training pipeline, massive compute and access to expensive human feedback or a powerful AI model to evaluate every response,” notes AWS.
RFT on Amazon Bedrock simplifies the model customisation process, opening the technique to any developer at any organisation.
It’s simples
The process, promises AWS, is simple.
“Developers select their base model, point it at their invocation logs (in other words, the AI’s history), or upload a dataset. Then, they choose a reward function—AI-based, rule-based, or a ready-to-use template. Automated workflows in Amazon Bedrock handle the fine-tuning process end-to-end. No PhD in machine learning required—only a clear sense of what good results look like for the business. At launch, RFT in Amazon Bedrock will support the Amazon Nova 2 Lite model. Compatibility with additional models is coming soon,” details AWS.
Developers can access advanced customisation techniques like Reinforcement Learning from AI feedback, Reinforcement Learning with Verifiable Rewards, Supervised Fine-Tuning and Direct Preference Optimisation.
The new SageMaker AI capabilities will work with Amazon Nova and popular open weight models like Llama, Qwen, DeepSeek and GPT-OSS, giving customers a wide range of options to match the right model to their use case.
