- Generative AI›
- Amazon Bedrock›
- Customize
Customize your applications
Securely tailor generative AI applications and agents to improve accuracy and relevance
Build secure, tailored AI applications with enterprise data
Organizations can leverage their unique enterprise data to build differentiated experiences for their business. Using techniques such as Retrieval Augmented Generation (RAG), model fine tuning, model distillation, and multimodal data processing, you can build generative AI applications tailored to your specific use case. Maintain complete control over sensitive information—your data is never used to train base models or shared with any model providers, including Amazon.
Make AI uniquely yours
Combine multiple customization tools to train models on your data and maximize model performance and accuracy for your unique requirements
Amazon Bedrock Knowledge Bases
Amazon Bedrock Knowledge Bases offers an end-to-end managed RAG workflow that lets you create highly accurate, low-latency, secure, and custom generative AI applications by incorporating contextual information from your own data sources.
- End-to-end RAG workflows
- Securely connect FMs and agents to data sources
- Deliver accurate responses at runtime
Supervised fine-tuning
Train a foundation model on your own data to improve performance on specific tasks. Fine-tuning teaches models your unique terminology, writing style, output formats, and domain-specific knowledge—making them more accurate and consistent
for your use cases.
Use fine-tuning when you need models to:
- Follow specific output structures or formatting requirements
- Master specialized vocabulary, technical language, or internal terminology
- Mimic a particular tone, voice, or brand style
- Improve accuracy on repetitive, well-defined tasks
Reinforcement fine-tuning
Reinforcement fine-tuning improves model accuracy by using feedback on responses rather than large datasets. You teach models what "good" looks like by rating their outputs, allowing them to learn your preferences and quality standards through iterative feedback. Use reinforcement fine-tuning when you have clear success criteria but limited training data, or when your requirements evolve over time. Reinforcement fine-tuning in Amazon Bedrock delivers 66% accuracy gains on average over base models.
Data automation
Amazon Bedrock Data Automation is a fully managed API that can easily integrate into your applications. It streamlines the development of generative AI applications and automates workflows involving documents, images, audio, and videos.
- Build intelligent document processing, media analysis, and other multimodal data-centric automation solutions
- Industry-leading accuracy at lower cost along with features such as visual grounding with confidence scores for explainability and built-in hallucination mitigation
- Integrated with Amazon Bedrock Knowledge Bases, making it easier to generate meaningful information from unstructured multimodal content to provide more relevant responses for RAG
Model distillation
With Amazon Bedrock Model Distillation, you can use smaller, faster, more cost-effective models that deliver use case–specific accuracy—comparable to the most advanced models in Amazon Bedrock. Distilled models in Amazon Bedrock are up to 500% faster and up to 75% less expensive than original models, with less than 2% accuracy loss for use cases like RAG.
- Fine tune a ‘student’ model with a ‘teacher’ model that has the accuracy you want
- Maximize distilled model performance with proprietary data synthesis
- Reduce cost by bringing your production data. Model Distillation lets you provide prompts and then uses them to generate responses and fine-tune the student models
- Boost function calling prediction accuracy for agents. Enable smaller models to predict function calling accurately to help deliver substantially faster response times and lower operational costs
Did you find what you were looking for today?
Let us know so we can improve the quality of the content on our pages