Amazon Bedrock: AWS GenAI for All | No Code-100% Drag & Drop

Explore bedrock with party rock, building a restaurant app using no-code drag-and-drop tools and country-based menus, powered by titan, llama, and stability ai.

Explore Bedrock's user interface, including the playground, Bedrock Studio, and builder tools, to access foundation models and text or image generation in a no-code drag-and-drop workflow.

Enable model access in AWS Bedrock by selecting all or specific models and providing organization details. Submit to grant access; some models require payment setup and cross-region reference.

Explore post model access in Amazon Bedrock, see the interface after access is granted, and note that ai21 labs, anthropic, llama, and meta models may require AWS team approval.

Explore the chat playground to test multiple models, compare providers, adjust prompts and metrics, and select the best Amazon model for chat utilization.

Explore how prompts modify responses from foundation models in a chat setting, using llama and prompt examples to change outputs, accuracy, and tone.

Explore the text playground in Amazon Bedrock, a single input-output tool, and compare models like I21, GPT, Gemini, Mistral AI, and llama.

Explore the image playground to generate text-to-image outputs with Amazon Titan and Stability AI, comparing speed and image quality while producing tiger and zoo scenes.

Explore what artificial intelligence is, its key components like machine learning, NLP, and computer vision, and how AI shapes healthcare, finance, and daily life.

Discover opportunities of generative ai across healthcare, education, creative arts, and business automation, while addressing ethical challenges like deepfakes, misinformation, bias, and intellectual property ownership.

Define artificial intelligence and its goal to simulate human tasks, and summarize machine learning, deep learning, and generative AI, including supervised, unsupervised, reinforcement learning, and convolutional neural networks.

Understand foundational models and large language models, their differences, and applications across text, images, and videos, with emphasis on transformer architecture, adaptability, and context-driven use cases.

Explore how retrieval augmented generation combines information retrieval with text generation to produce up-to-date, context-aware responses from a knowledge base, supporting customer support, knowledge management, and healthcare.

Explore how crawling and indexing power data retrieval, with web crawlers (spiders) collecting pages, scheduling updates, and building fast, organized indexes for quick, accurate searches.

Design and refine prompts to guide foundation models with clear instructions, context settings, and persona, delivering relevant and accurate responses through iterative testing and examples.

Explore the four basic prompt types for foundation models: instruction based prompts, few short learning prompts with examples, chain of thoughts, and persona prompts, and learn how each guides responses.

Explore Bedrock pricing, including pay-as-you-go usage, model inference, fine tuning, storage and retrieval costs, with token-based calculations and OpenSearch as a major cost driver.

Explore inference parameters that control randomness and quality in foundation model text generation. Learn how temperature, nuclear sampling with top p, and top k shape diversity, predictability, and response style.

Discover how temperature ranges from 0 to 1 govern creativity and predictability in generative AI, with lower values yielding predictable outputs and higher values boosting storytelling creativity.

Explore how adjusting the temperature setting affects model creativity and randomness with a second hands-on example using the anthropic chat model, showing predictable versus creative outputs.

Explore how the top parameter, ranging from 0 to 1, balances predictability and creativity with temperature. See practical model settings and end-to-end generation examples.

Explore top-p sampling hands-on with the cloud three sonnet model, comparing predictive and creative outputs from a simple example like the sound of a dog.

Explore how configuration, temperature, and top-p shape generative model outputs, showing they can be predictable, creative, or random, with real-time dog-sound examples illustrating a balanced approach.

Learn to increase the chances of getting the desired answers by tuning temperature and top-p, and using their combination to balance predictability and creativity in generative AI.

Learn how top k constrains responses by selecting the top results from model outputs, balancing randomness and predictability with temperature, probability, and model availability.

Learn to tune temperature, top p, and top k to guide generative ai toward the most appropriate result with the anthropic sonnet model, noting randomness.

Explore how additional configuration parameters tune model outputs, including temperature, top P, and top k, and use system prompts, max length, and stop sequences to control output.

Explore system prompts as personas that guide AI behavior, optionally used across use cases. See examples where a model acts as a kid or a smart kid.

Learn how to set the maximum length for outputs by selecting a model and adjusting word or token limits. See how length impacts completeness and cost with a 200-word example.

Explore how stop sequences end model outputs when a pattern or keyword is matched, with real-time demonstrations using AI21 labs models and default 0.5 settings.

Understand safeguards for generative AI, including guardrails to block harmful topics. Learn how watermark detection identifies AI-generated images to support safe, ethical use, content filtering, access control, and compliance.

Set up guardrails to restrict sensitive topics, name and configure guardrails, specify block messages, enable KMS encryption, block PII, and test across models to ensure policy compliance.

Apply guardrails to chat interactions with the anthropic sonnet model, upgrade versions as policies evolve, and manage working drafts to align behavior with industry norms.

Explore watermark detection in Amazon Bedrock by testing images from Titan and Stability AI, showing that watermark detection works with Titan-generated images and not with Stability AI.

Master builder tools for AWS GenAI, including prompt management, version control, templates, and agents. Leverage knowledge base and prompt flow to tailor foundation model interactions and streamline tasks.

Master prompt management with the builder tool, creating and testing prompts and variables for industry use, then compare models like Titan and Sonnet for manufacturing AI.

Create a knowledge base in Amazon Bedrock by uploading files to an S3 bucket, selecting data sources, and configuring embedding models to tailor answers for your organization.

Learn about IAM users and IAM roles in AWS Bedrock, highlighting long-term versus temporary access, policies, and secure service integration with examples like EC2 and S3.

Explore how IAM policies define permissions for users, groups, and roles to access AWS resources using JSON documents, with statements that cover actions, resources, conditions, and custom policies you attach.

Create an IAM user and grant administrator access to set up the bedrock knowledge base. Sync data sources, build the vector database with embeddings, and query insights from the data.

Explore how Amazon Bedrock agents act as task-specific helpers, executing natural language instructions to perform data retrieval, content generation, and customer interactions with integrated APIs, data sources, and automated workflows.

Learn to create agents that act as assistants for hotel or flight bookings, using manual or assisted workflows. Configure models (Anthropic, Amazon, Meta), knowledge bases, memory, and guardrails, then test.

Create and configure agents with the assistant, automating tasks using prompts and action groups. Prepare, test, and run agents that can book flights, hotels, and rental cars with configurable parameters.

Define HR-based agents manually using knowledge bases, memory, and guardrails, configure encryption and prompts, save steps, and test policy queries end-to-end.

Explore building end-to-end agents by adding action groups and guardrails, configuring user input for date and days, and using parameters to drive Lambda functions via API gateway.

Explore prompt flow in Amazon Bedrock as a drag-and-drop, no-code pipeline that orchestrates prompts, data fetching, and outputs to tailor generative AI responses.

Learn to build end-to-end workflows with prompt flow using drag-and-drop, integrating knowledge bases, permissions, KMS encryption, and lambda-powered code, while testing and troubleshooting outputs.

Utilize the prompt flow builder to use prompts, configure them with topic and industry, and run outputs for artificial intelligence in manufacturing using the sonnet model.

Create and test end-to-end prompt flows for a music playlist use case. Configure prompts, models, and inference settings, then verify knowledge base utilization and HR-style prompt outputs.

Explore the condition builder in promptflow by creating if-else flows that evaluate cloud input, use operators like equal to, not equal to, and route prompts to AWS or non-AWS prompts.

Build a number input prompt flow to check voting eligibility by age, using greater than or equal to 18 conditions, multiple prompts, and end-to-end testing of eligible or ineligible responses.

Explore Bedrock Studio, a fully managed drag-and-drop console and IDE to build, customize, and deploy generative AI apps with prompt flows and pre-trained models.

Master centralized management of multiple AWS accounts using IAM organization to centrally control policies, service access, consolidated billing, and organizational units.

Enable and configure the AWS IAM Identity Center for Bedrock Studio by linking an AWS Organization, creating accounts and permission sets, and managing users and MFA.

Learn how to add users in IAM Identity Center, assign permission sets, and grant Bedrock Studio access with MFA and administrator rights for foundation model usage.

Enable and verify an AWS Bedrock user via IAM identity center, send a verification link for email-based two-step verification and MFA, then manage end-to-end permission sets and account access.

Set up Bedrock Studio in the AWS console, configure permission boundaries, create inline policies and roles, and establish trust relationships to enable data zone access for Bedrock.

Create and configure a Bedrock Studio workspace using IAM identities, provisioning roles, and model choices, then manage members, tags, encryption, and monitoring for your utilization.

Assign members to the workspace by searching for the created DevOps agents, then confirm and keep adding users as needed, using groups from the IAM Identity Center for access.

Explore the Bedrock Studio workspace UI in the AWS console, sign in with IAM, manage users, and build and explore projects using chat and prompt flow.

Build a basic chat app in AWS Bedrock Studio, configure a movie expert system prompt, adjust temperature, and test generated top ten movie suggestions.

Learn to apply guardrails to your chat app using AWS Bedrock, creating deny topics like horror movies, configuring content filters, and testing the guardrails in the console.

Explore editing and deleting guardrails in a chat app, modify scenarios, and test changes within the app to ensure guardrails apply or are removed as needed.

Learn to integrate knowledge bases into a chat app by uploading KB pages with item embeddings and OpenSearch as vector database, and address lambda memory size issues with AWS support.

Resolve a size error in the chat app by requesting AWS support to increase the Lambda function size, then build and utilize a knowledge base to generate answers in chat.

Discover how project isolation keeps components separate, preventing cross-project reuse; see how a default, test, and devops project keep their KB pages and apps isolated, with cross-project components unavailable.

Add prompts to the prompt flow to generate a movie list by genre using a prompt node, knowledge base, and Bedrock Studio, then test end-to-end.

Design a conditional prompt flow that blocks Bollywood requests and directs users to alternative genres, using versioned prompts and a knowledge-base fallback in Amazon Connect.

Select the right foundation model to optimize cost and accuracy for your use case. Size, data relevance, and task needs shape outputs.

Identify criteria for selecting the right fm for a use case, including data relevance, performance metrics, model size, cost, and integration with existing workflows.

Explore three model evaluation options—automatic, AWS managed, and bring-your-own-team—using built-in or your own datasets, and inspect metrics like accuracy and toxicity across text generation, summarization, Q&A, and classification.

Explore bedrock models for natural language understanding, text generation, multilingual writing, and image generation, including Anthropic, AI21 Jurassic-2, Titan Text, Stability AI Stable Diffusion, and Cohort for retrieval augmented generation.