DEEPCHECKS GLOSSARY

AWS Bedrock

What is Amazon Bedrock?

Amazon Bedrock is a cutting-edge, fully managed service offered by Amazon Web Services (AWS) that provides developers access to foundational models (FMs) for building artificial intelligence (AI) applications. Bedrock offers a selection of high-performing FMs from leading AI companies such as Anthropic, Cohere, Meta, Mistral AI, Stability AI, and AI21 Labs, all accessible through a single API. Additionally, Bedrock provides tools to ensure the responsible use of AI, with features designed to ensure security and privacy.

Amazon Bedrock is serverless and integrates well with many existing AWS services that applications might already be using. This makes it ideal for securely and efficiently deploying generative AI capabilities into existing customer solutions.

Amazon also recently introduced the capability to bring your AI models into Bedrock. The ability to comparatively evaluate various foundational models alongside your own to select the best one for your use case makes Bedrock an attractive solution for building generative AI applications.

Key features provided by Amazon Bedrock

  • Wide range of foundation models: Amazon Bedrock provides access to a variety of foundational models with multi-modal capabilities. Notable models include the Jurassic-2 model for multi-lingual text generation, Amazon Titan for text summarization and classification, Claude and Llama for content generation and question-answering, and Stable Diffusion for image generation.
  • Serverless integration: Amazon Bedrock is serverless and fully managed, simplifying the deployment and scaling of generative AI models. This significantly reduces operational overhead, allowing developers to focus on building applications.
  • Model customization: Fine-tuning features provided by Amazon Bedrock allow customers to safely and efficiently customize and adapt models with their proprietary data. Bedrock creates private copies of the model, which can be fine-tuned using the retrieval augmented generation (RAG) technique. This enables the models to understand the context of their organization better and generate more informed, accurate responses.
  • Security, privacy, and safety: Amazon manages the models provided by their vendors within their ecosystem, ensuring that customers’ proprietary data stays within AWS and is encrypted both in transit and at rest. Users can leverage existing security controls such as KMS for encryption, CloudTrail for governance, and VPC for network security to ensure maximum safety. Furthermore, Amazon complies with standards such as ISO, SOC, HIPAA, CSA, and GDPR, making it suitable for use in sensitive and regulated industries.
  • Agents for Amazon Bedrock: Agents for Bedrock help automate the steps required to integrate generative AI applications with complex, multi-step tasks across company systems and data sources. Agents streamline applications with customized instructions, orchestration, and enhanced monitoring. For example, complex tasks such as creating ad campaigns and processing insurance claims can be automated without compromising security and privacy.
  • Guardrails for Amazon Bedrock: Guardrails for Bedrock helps implement safeguard policies on generative AI applications to help ensure the responsible use of AI. This acts as a layer of security on top of the underlying models and facilitates blocking undesirable topics, filtering harmful content, and redacting sensitive information (Personally identifiable information).
Deepchecks For LLM VALIDATION

AWS Bedrock

  • Reduce Risk
  • Simplify Compliance
  • Gain Visibility
  • Version Comparison
TRY LLM VALIDATION

Use cases of Amazon Bedrock

  • Text generation: Text generation with generative AI is one of its most lucrative capabilities. By leveraging the foundational models provided by Amazon Bedrock, generating textual content has never been easier. Text for stories, social media posts, blog posts, news articles, and much more can be generated, which is particularly useful in industries such as marketing, entertainment, personalization, and gaming.
  • Virtual assistants: Amazon Bedrock empowers customers to build chatbots or virtual assistants trained on proprietary data. This results in virtual assistants that can better understand user requests, engage in natural conversation to collect more information, automatically break down tasks to fulfill those requests, and perform the necessary actions.
  • Image generation: Through foundational models such as Stable Diffusion and Amazon Titan, image generation is quite straightforward in Amazon Bedrock. It can generate photorealistic, unique, artistic images responding to natural language queries, which is immensely helpful in building websites, ad campaigns, and presentations.
  • Text summarization: Through the power of foundational models, Amazon Bedrock can generate concise summaries of long documents such as books, reports, articles, research papers, and technical documents. This allows for quick and efficient information extraction, significantly boosting productivity.
  • Text and image search: Generative AI-powered search can synthesize and provide relevant information to user queries based on large datasets of textual and image content. Foundational models in Amazon Bedrock help streamline workflows that require searching, making the process seamless and efficient.