Amazon is quickly emerging as one of the biggest artificial intelligence companies in the world today. For some time now, the company has offered businesses access to innovative AI solutions in a range of software systems and applications available through AWS (Amazon Web Services). Amazon Bedrock Studio represents Amazon’s attempt to streamline generative AI app generation for teams.
Introduced in “Preview” mode in May 2024, Amazon Bedrock Studio is an SSO-enabled web interface that provides users with easy access to tools for developing and optimizing AI tools. The Studio enables direct access to a range of foundational models, including the new selection of “Nova” models introduced by Amazon in late 2024.
Plus, it allows developers to use their company credentials and data to build, evaluate, prototype, and share generative AI apps. Here’s everything you need to know about Amazon Bedrock Studio.
What is Amazon Bedrock Studio?
Amazon Bedrock is the fully managed service that offers companies access to a range of high-performing foundation models they can access for various use cases. The full service provides direct access to numerous high-performing foundation models from companies like Meta, Mistral AI, Anthropic, A121 Labs, and of course, AWS.
Using Amazon Bedrock, users can easily experiment with various foundational models for different use cases, and customize them with their data using fine-tuning and Retrieval Augmented Generation (RAG) techniques. Amazon Bedrock Studio is the SSO-enabled interface that streamlines building AI applications with Bedrock capabilities.
This “Studio” is basically the AWS alternative to Google’s Cloud Vertex AI agent builder, and Azure’s Copilot Studio. It’s a low-code/no-code, assistant builder, where developers can add documents as data sources, and automatically vectorize them to support RAG. They can also take advantage of prompt engineering techniques, and implement guardrails for AI governance.
Amazon Bedrock IDE is also integrated with the Amazon SageMaker Unified Studio, to support easy collaboration across teams. Developers and technical stakeholders can work together to collectively build and customize AI applications, and make granular changes to each system.
The Key Features of Amazon Bedrock Studio
With Amazon Bedrock Studio, AWS administrators can create workspaces for their organization within the Bedrock Management Console. They’ll be able to grant specific permissions to individuals and groups, allowing developers to log in using their SSO credentials and start working with foundation models and other tools in a “sandbox” space.
Although similar AI solutions exist, like Microsoft’s Studio built in collaboration with OpenAI, Amazon Bedrock Studio offers a handful of unique features and benefits.
The key features of the platform include:
An AI Playground with Dozens of AI Models
Within the Amazon Bedrock Studio playground, developers can experiment with and test various models and configurations. As mentioned above, there are dozens of large language models (LLMs) and foundational models to choose from, such as Meta’s Llama 3 series and Amazon Nova models.
Notably though, some models do have restrictions. For instance, if you want to build something with an Anthropic model, you need to submit details of your use cases in advance. Plus, it’s worth noting that you do need access to a configured AWS workspace with pre-set SSO and roles.
Bedrock Knowledge Bases
The Knowledge Bases in Amazon Bedrock offer a solution for out-of-the-box Retrieval-Augmented Generation (RAG). For those unfamiliar, RAG is a technique used by AI models which involves retrieving data from specific sources. With knowledge bases in Amazon Bedrock, users can take advantage of this functionality without the need for extra configuration.
All you need to do is upload documents and data, which are automatically ingested by the Amazon Titan model, and stored in the Amazon OpenSearch Serverless instance. The real benefit of these Knowledge Bases is their ability to augment the prompts you give to your foundational models with more contextual information from uploaded files. This can help to lead to more relevant and accurate responses from bots designed for specific purposes.
Plus, responses informed by data from your knowledge bases include citations o you can track the source of the information and verify it for factual accuracy.
Guardrails for AI Governance
For those concerned about safe and ethical AI, guardrails in Amazon Bedrock Studio empower creators to implement safeguards for generative AI applications based on certain use cases and responsible policies. Users can create a range of guardrails tailored to specific scenario.
For instance, you can use guardrails to define “denied topics”, which stop generative AI solutions from responding to questions about dangerous subjects. You can use filters to block harmful content in both user inputs and model responses. Plus, you can create content filters, and use sensitive information filters to block AI from accessing crucial data.
For now, guardrails are only available for use with text-only foundation models. However, future iterations of Amazon Bedrock Studio might support similar solutions for multimodal apps.
Amazon Bedrock Studio Functions
Function calling is a valuable and crucial feature that allows models to call functions to access certain modules and capabilities when handling a prompt. This basically means you can infuse generative AI models with “fresh” information that wasn’t included in their training or knowledge bases.
With function calling, users can incorporate APIs to pull updated information into a response. For instance, you could allow the AI model to browse the web to find out about the latest weather forecasts, or sport results.
Functions basically extend and improve the capabilities of generative AI models by allowing them to access and dynamically incorporate more valuable information into each response. Even more compelling, is that with Amazon Bedrock Studio, users don’t need to write code to use APIs.
AWS handles the hard work for you, creating Lambda functions automatically after users provide an OpenAPI schema to the system.
The Benefits of Amazon Bedrock Studio
Amazon’s Bedrock Studio solution is still in its early stages, but it has a lot of value to offer. Compared to some other toolkits available for creating generative AI apps, for instance, Bedrock Studio is a lot more cost effective. There aren’t any complex fees tied to the platform.
You only pay for the Bedrock capabilities you use, such as API calls to knowledge bases and foundational models. Beyond that, Amazon Bedrock Studio benefits from:
Access to Versatile Foundation Models
Amazon Bedrock doesn’t just offer access to Amazon’s own AI and foundation models. It allows creators to take advantage of the latest innovations in the market. There are a huge range of foundation models to choose from already, from companies like Anthropic, Cohere, AI21 Labs, and Stability AI. Plus, Amazon Bedrock has its own marketplace where you can discover, test, and experiment with various emerging FMs in fully managed endpoints.
With the single API access structure of Amazon Bedrock, regardless of which model you choose, you’ll be able to flex between different systems, and upgrade to the latest model versions over time, without investing in complex code changes. You can even import your own custom models.
Extensive AI Fine-Tuning
With Amazon Bedrock Studio, companies can create highly customized generative AI models for specific use cases. Users can privately fine-tune the models they access using their own labeled data sets with minimal upfront effort. You also get access to continued pre-training, and Amazon Bedrock makes a separate copy of the foundational model only businesses can access, to improve privacy.
Beyond this, as mentioned above, you can equip your foundation models with up-to-date information using “RAG”. The Knowledge Bases in Bedrock Studio give you access to RAG capabilities that allow you to adapt responses with company data. The Knowledge Bases automate the RAG workflow, from ingestion and retrieval, to prompt augmentation and citations. This removes the need to create custom code for source data integration.
Additionally, for unstructured data sources with multimodal data, users can configure the Knowledge Bases in Bedrock to parse and extract meaningful insights. There are options to choose between Bedrock Data Automation and Foundation models for parsing.
Multi-Stage Tasks (Agentic AI)
Amazon Bedrock Studio also enables access to agentic AI capabilities, allowing companies to leverage machine learning and AI to create bots capable of multi-stage tasks. For instance, you can create a bot that can answer customer questions by tapping into your existing software to check product availability, inventory levels, and order information.
Bedrock Studio makes it simple to create agents in just a few steps. All users need to do is select the foundation model that they want to use, and give that model access to their enterprise systems through APIs, and knowledge bases. You’ll also need to enable access to AWS Lambda functions to allow the model to securely execute APIs too.
Created agents can analyze user requests and automatically call on the right APIs and data sources to complete a task. Plus, these agents in Bedrock come with built-in security and privacy features, to help enhance AI governance.
Versatility for Numerous Tasks
Already there are numerous ways to leverage Amazon Bedrock Studio to create agents capable of various workplace tasks. For instance, with Amazon Bedrock Data Automation, companies can streamline the generation of insights from unstructured multimodal content like images and audio.
This allows companies to build generative AI applications that can automat all kinds of workflows, from media analysis, to research, in a cost-effective manner. In the months and years ahead, Amazon will be releasing new capabilities for the Amazon Bedrock ecosystem too.
For instance, users will be able to take advantage of “Luma AI” to generate video clips from text images – similar to what you can do with tools like OpenAI Sora and Google Veo.
Poolside, another tool specifically designed for software development workflows, will also be coming soon, to help companies create AI solutions dedicated to software engineering tasks.
Getting Started with Amazon Bedrock Studio
Notably, Amazon Bedrock Studio still isn’t available to all users worldwide. AWS is gradually rolling capabilities out to different groups in specific countries, so you’ll need to visit the Amazon Bedrock website for insights into whether you can actually access the system.
If you do have access to the studio, the first thing you’ll need to get started is an AWS Administrator account. Using that account, you’ll have to create a dedicated “Amazon Bedrock Studio” workspace, and select the users you want to add to the space. You can choose to give different users varying permissions within the workspace.
Creating and Using your Workspace
To create your workspace, navigate to the Amazon Bedrock Console and choose the “Bedrock Studio” option on the bottom left-hand side. You’ll be asked to configure and secure the single sign-on integration with your chosen identity provider in the AWS IAM Identity Center.
Once you’ve done that, choose the “Create Workspace” button, and enter the details requested. At this point, you’ll be asked to create specific AWS IAM roles for users. You’ll also be able to select the default generative AI and embedding models you want to use within the workspace. You can choose to define these later if you prefer.
Once you’re done, hit the Create button, and select the workspace you’ve created. Click on the User Management button, and select Add Users or Groups to pick the people you want to add to your Workspace. Once you’ve selected your users, you can head back to the Overview tab and copy the URL to the Amazon Bedrock Studio environment to share with your users.
Once users get access to this URL, they can sign in using their SSO credentials, and start experimenting with the various models you’ve selected, or the available models in the platform. They can integrate their own data, leverage Knowledge Bases, experiment with guardrails, and use functions to make API calls.
Plus, once they create a generative AI app, they’ll be able to share it with their team members and invite them to help fine-tune the model.
Simplifying AI Development with Amazon Bedrock Studio
The use cases for Amazon Bedrock Studio are growing all the time. On its website, AWS draws attention to the opportunities companies have to create generative AI agents that can generate text, conduct text and image searches, and even produce high-quality images.
You can even create agents that can summarize text for you, or produce comprehensive “agents” that perform multi-stage tasks related to supply chain management or customer service.
Various companies have already begun experimenting with the technology. For instance, Toyota is using Amazon Bedrock to power its “Safety Connect” project. The Alida brand is using Amazon Bedrock and the Claude foundational model to gain deeper insights into customer feedback.
Lonely Planet has also used Bedrock to reduce itinerary generation costs by 80%, and streamline various workplace processes. Although the capabilities offered by Amazon Bedrock Studio might not be entirely new (they were already available to a certain extent through Amazon Bedrock), the studio solution introduces a new opportunity to developers.
The seamless playground environment combines all of the features creators need to design and experiment with AI models into a single ecosystem. This significantly lowers the barrier to entry for generative AI development, and could power a future of rapid iteration and discovery.
If you want to put the solution to the test yourself, you can start experimenting with Amazon Bedrock Studio in certain regions worldwide with an AWS Bedrock account today.
