AWS Infrastructure and Services for Building GenAI Applications

Amazon Bedrock is the flagship AWS service for generative AI, and the one the exam expects you to reach for first. It is a fully managed, serverless service that gives you access to a choice of high-performing foundation models from Amazon and leading AI companies through a single API. Providers available through Bedrock include Amazon (the Nova models), Anthropic, Meta, Mistral AI, Cohere, AI21 Labs, and Stability AI.

Because Bedrock is serverless, you do not provision or manage any servers, GPUs, or clusters. You send a prompt to a model and get a response back. You can also switch between models by changing a model identifier, which makes it easy to compare models or move to a better one later.

Bedrock also bundles higher-level building blocks so you do not have to assemble them yourself. Knowledge Bases connect a model to your own data so it can answer using that data (retrieval augmented generation, or RAG). Agents let a model carry out multi-step tasks by calling APIs and tools. Guardrails apply safety and content filters to inputs and outputs. You will study each of these in depth in later domains; for this task, simply recognize that they are Bedrock features and that Bedrock is the fast, managed path to using foundation models.

Amazon SageMaker AI is the AWS platform for building, training, and deploying machine learning models, and it is the service you choose when you need more control than Bedrock gives you. With SageMaker AI you can bring your own data and algorithms, train a model from scratch, fine-tune an existing model, and host the result on endpoints you configure. That control comes with more responsibility: you make more decisions about instances, training jobs, and deployment than you do with a serverless API.

Amazon SageMaker JumpStart is a machine learning hub inside SageMaker AI that shortens the path. It offers pre-trained, publicly available foundation models and built-in solutions you can deploy with a few clicks, then fine-tune on your own data. JumpStart is a useful middle ground: you get ready-made models like those on Bedrock, but you deploy and host them yourself within SageMaker, keeping full control over the environment.

A simple way to remember the split: reach for Bedrock when you want to consume foundation models quickly through a managed API, and reach for SageMaker AI when you want to build, train, fine-tune, or host models yourself. Many teams use both, running Bedrock for fast features and SageMaker AI for custom model work.

Beyond the two core services, the exam expects you to recognize a handful of supporting technologies by name and purpose.

Amazon Q is a generative AI-powered assistant. Amazon Q Business answers questions and generates content grounded in your company's own data and systems, acting as an AI assistant for the business. Amazon Q Developer helps software teams by explaining code, suggesting changes, and answering development questions. When a question describes an AI assistant for employees or developers rather than a model you call from code, Amazon Q is the answer.

Kiro is an AI-assisted development tool, an agentic environment that helps developers plan and build software with AI. Recognize it as a developer productivity tool.

Strands Agents and Amazon Bedrock AgentCore are the newer agent-building technologies. Strands Agents is an open-source toolkit for building AI agents in code. Amazon Bedrock AgentCore is a set of services for deploying and operating AI agents securely at scale, providing the runtime and supporting infrastructure agents run on. For this foundational task you only need the one-line definition of each: they help you build and run AI agents. The deeper mechanics of agents belong to later domains.

Use this table to fix each service to the single job it does. On a recognition-style question, the wording of the scenario usually points straight at one row.

A quick test-day check: if the scenario is about calling a model, think Bedrock; if it is about building your own model, think SageMaker AI; if it is about an assistant that helps people, think Amazon Q.

AWS positions its managed generative AI services around a set of advantages you should be able to list. Accessibility and a lower barrier to entry come first: a managed API means you can start using powerful foundation models without deep machine learning expertise, a data science team, or your own GPUs. Efficiency follows, because AWS runs the undifferentiated heavy lifting of provisioning, scaling, patching, and operating the model infrastructure, so your team spends its time on the application rather than the plumbing.

Cost-effectiveness comes from paying only for what you use instead of buying and running hardware that sits idle, and from being able to choose a smaller, cheaper model when it is good enough. Speed to market is a major draw: because the models and building blocks are ready to call, you can move from idea to a working prototype in days rather than months.

Together these let you meet business objectives faster. You can test whether generative AI actually solves a problem, then scale it up if it does, without a large upfront commitment. In short, managed services trade some control for speed, simplicity, and lower risk.

Choosing AWS for generative AI is not only about the models; it is about the infrastructure they run on. The exam highlights four benefits.

Security: your data is protected with encryption in transit and at rest, and you control access with AWS Identity and Access Management (IAM). Importantly, with Amazon Bedrock your prompts and data are not used to train the base foundation models, and your content stays within your control.

Compliance: AWS services operate within a wide range of compliance programs and certifications, which helps you meet regulatory obligations in regulated industries without building that foundation yourself.

Responsibility: AWS provides tools that support responsible AI, such as Amazon Bedrock Guardrails for filtering harmful content and blocking unwanted topics, so you can put safeguards around model behavior.

Safety: the AWS shared responsibility model, private networking options, and monitoring let you run generative AI within your own secure environment rather than sending data to an unmanaged third party.

You will study security and governance in depth in a later domain. Here, simply connect each benefit (security, compliance, responsibility, and safety) to the idea that AWS gives you a trusted foundation to build generative AI on.

Running generative AI on AWS involves cost tradeoffs you should recognize. Most Bedrock usage is priced per token: you pay for the tokens in your input (the prompt) and the tokens the model generates (the output). Because larger prompts and longer answers cost more, prompt and response length directly affect cost. Larger, more capable models generally cost more per token but may give better answers, so you balance performance against price.

AWS offers two main ways to pay for model inference. On-demand is pay-as-you-go per token with no commitment, ideal for variable, low-volume, spiky, or experimental workloads. Provisioned Throughput lets you reserve a fixed amount of model capacity for a set hourly rate over a time commitment, giving predictable, high-volume throughput at a lower effective cost per request, ideal when traffic is steady and heavy. Provisioned Throughput is also required to run some custom (fine-tuned) models.

Other levers affect cost and experience: responsiveness (latency), availability and redundancy across Availability Zones, and regional coverage, since a model may only be available in certain AWS Regions and running closer to your users can reduce latency. Match the pricing model to your traffic pattern: on-demand for unpredictable or small workloads, Provisioned Throughput for predictable high volume.

When a question asks which service fits, compare them on control and effort.

Scenario: a retail company wants to add a customer-service chatbot to its website in the next few weeks. It has no machine learning team and wants to answer questions using its existing help-center documents. Amazon Bedrock is the right choice: the team can call a foundation model through the API and use Knowledge Bases to ground answers in the help-center content, shipping quickly with minimal expertise. Later, the same company needs a specialized model trained on its own proprietary product data, with full control over training and hosting to meet strict internal requirements. That is a job for Amazon SageMaker AI. Many organizations use Bedrock first for speed, then adopt SageMaker AI where a custom model earns its extra effort.