This course introduces Amazon Bedrock and its use in building secure, scalable generative AI applications on AWS. Participants learn to select foundation models, design prompts, integrate model APIs, connect organizational data through retrieval-augmented generation (RAG), and configure AI agents and guardrails. The course also covers security, monitoring, and cost optimization, concluding with a capstone that combines these capabilities into a working knowledge assistant.

 

Duration 5 Days – 35 hrs. 

 

Objectives

• Explain Amazon Bedrock capabilities and common generative AI use cases.
• Select foundation models based on application requirements, capabilities, latency, and cost.
• Configure AWS permissions and development resources for Amazon Bedrock.
• Design prompts and adjust inference settings to improve model responses.
• Integrate foundation models into applications using Python and Amazon Bedrock APIs.
• Build RAG applications using Amazon Bedrock Knowledge Bases.
• Configure Amazon Bedrock Agents to retrieve information and invoke application functions.
• Apply Amazon Bedrock Guardrails and appropriate security controls.
• Monitor application behavior and manage performance and operating costs.
• Build an integrated generative AI knowledge assistant.

 

Target Audience 

• Software developers and application engineers.
• Cloud engineers and AWS solution architects.
• AI and machine learning engineers.
• Data engineers and data scientists developing generative AI solutions.
• Technical leads and technology specialists responsible for AI adoption.

 

Prerequisites

• Basic understanding of AWS services and cloud computing.
• Familiarity with AWS Identity and Access Management (IAM) and Amazon S3.
• Working knowledge of Python programming.
• Basic understanding of APIs, JSON, and client-server applications.
• Introductory knowledge of AI and machine learning concepts; advanced model training experience is not required.
• Access to an AWS training account with appropriate permissions, available foundation models, and sufficient service quotas in the selected AWS Region. 

Course Outline 

Day 1: Amazon Bedrock Foundations and Environment Setup

Module 1: Introduction to Generative AI and Amazon Bedrock

• Generative AI concepts and business applications.
• Foundation models, large language models, and multimodal models.
• Tokens, embeddings, context windows, and inference.
• Amazon Bedrock capabilities and its role in the AWS AI ecosystem.
• Model limitations, hallucinations, and responsible AI considerations.

 Module 2: Foundation Model Selection

• Exploring available model families and supported capabilities.
• Matching models to text generation, summarization, conversation, and other use cases.
• Understanding model and feature availability by AWS Region.
• Comparing response quality, latency, context capacity, and cost.
• Introduction to inference options and model customization approaches.

 Module 3: AWS Environment and Initial Model Invocation

• Configuring IAM identities, roles, and permissions.
• Confirming model access requirements and service quotas.
• Exploring models through the Amazon Bedrock console.
• Preparing a Python development environment and AWS SDK.
• Making an initial model request and interpreting the response.

 

Day 2: Prompt Engineering and Application Integration

Module 4: Prompt Design Fundamental

• Writing clear instructions and providing relevant context.
• Using system instructions and user messages.
• Applying zero-shot and few-shot prompting.
• Specifying output formats and response constraints.
• Managing ambiguous requests and insufficient information.
• Recognizing prompt injection risks.

 Module 5: Amazon Bedrock APIs and Inference Controls

• Using the Converse API with supported models.
• Understanding when to use model-specific invocation APIs.
• Managing conversation history and context.
• Configuring supported inference parameters.
• Implementing streaming responses.
• Handling throttling, retries, and service errors.

 Module 6: Building a Generative AI Application

• Designing a basic conversational application.
• Integrating Amazon Bedrock with Python application code.
• Implementing summarization and information extraction.
• Introduction to tool use and structured responses.
• Managing application configuration and credentials.

 

Day 3: Retrieval-Augmented Generation and Knowledge Bases

Module 7: RAG Architecture and Data Preparation

• Understanding retrieval-augmented generation.
• Comparing prompting, RAG, and model customization.
• Preparing organizational documents for retrieval.
• Understanding chunking, embeddings, and vector storage.
• Managing source quality, metadata, and data access.

 Module 8: Amazon Bedrock Knowledge Bases

• Configuring a knowledge base and supported data source.
• Selecting embedding models and compatible storage.
• Ingesting and synchronizing documents.
• Retrieving relevant content and generating grounded responses.
• Using source citations and metadata filters.
• Understanding retrieval and generation configuration options.

Module 9: Improving Knowledge-Based Responses

• Identifying retrieval gaps and irrelevant results.
• Refining chunking and retrieval settings.
• Introduction to reranking where supported.
• Handling missing, conflicting, and outdated information.
• Integrating a knowledge base into a conversational application.

  

Day 4: Agents, Guardrails, and Security

Module 10: Amazon Bedrock Agents

• Understanding agent instructions and orchestration.
• Creating an agent and selecting a compatible model.
• Connecting an agent to a knowledge base.
• Defining action groups and function parameters.
• Integrating AWS Lambda for application actions.
• Managing sessions and inspecting agent traces.
• Requiring confirmation for sensitive actions.

 Module 11: Amazon Bedrock Guardrails

• Understanding guardrail capabilities and limitations.
• Configuring content filters and denied topics.
• Applying word filters and sensitive information controls.
• Introducing contextual grounding checks.
• Integrating guardrails with model calls, knowledge bases, and agents.
• Designing appropriate blocked-response and fallback messages.

Module 12: Security and Governance

• Applying least-privilege IAM permissions.
• Protecting data through encryption and access controls.
• Understanding private connectivity options.
• Managing invocation logging and sensitive information.
• Using AWS CloudTrail for supported activity auditing.
• Addressing prompt injection and excessive agent permissions.

 

Day 5: Operations and Integrated Capstone

Module 13: Operational Readiness and Cost Management

• Monitoring service usage, latency, and errors with Amazon CloudWatch.
• Understanding token consumption and associated service charges.
• Managing quotas, budgets, and application usage limits.
• Optimizing model selection, prompt size, and response length.
• Introduction to prompt caching and batch inference where supported.
• Planning deployment, version management, and resource cleanup.

 Module 14: Capstone — Build an Organizational Knowledge Assistant

• Define a focused use case and application requirements.
• Prepare a sample organizational document collection.
• Configure a knowledge base for grounded question answering.
• Build a conversational interface using Python and Amazon Bedrock APIs.
• Configure an agent with a simple read-only lookup action.
• Apply guardrails and least-privilege permissions.
• Include source citations and responses for unavailable information.
• Add basic monitoring and cost controls.
• Document the solution architecture, configuration, and deployment considerations.