AWS Certified AI Practitioner: Hands-On + Practice Exams

Explore cloud computing fundamentals, including public, private, and hybrid clouds, and the three computing models (iaas, paas, saas), plus the key benefits and security of cloud services.

Understand cloud computing models: infrastructure as a service, platform as a service, and software as a service, with AWS examples like EC2, Elastic Beanstalk, and S3.

Learn how AWS infrastructure uses regions, availability zones, and edge locations to enable redundancy and low latency, with edge caches via CloudFront and Route 53.

Explore AWS concepts of highly available systems, fault tolerance, and disaster recovery. Understand regions, availability zones, autoscaling, load balancing, and backups to keep applications running.

Create custom AWS budgets to control cost and usage, and receive email alerts at 50%, 85%, and 100% thresholds.

Understand the root user and IAM user in AWS: the root has full unrestricted access, while IAM users have policy-based, day-to-day access managed by the root user.

Define permission sets for users, groups, and roles with IAM policies, attach them to grant access to AWS services, understand that policies are JSON documents with effect, actions, and resources.

Launch and configure elastic compute cloud (EC2) instances to run virtual servers with customizable cpu, memory, storage (EBS), networking, operating systems, regions, and architectures on AWS.

Discover how AWS Lambda enables serverless code execution and event-driven triggers. Scale automatically and support multiple runtimes for real-time data processing, web backends, and IoT apps.

Discover aws s3, the simple storage service, offering scalable object storage from anywhere with buckets that organize data, enforce unique 3–63 character names starting with a lowercase letter or number.

Understand Amazon VPC as a logically isolated boundary where EC2 and EBS reside, and learn to use subnets, route tables, internet gateway, and security groups for secure web hosting.

Explore how a real-time society demonstrates VPC concepts, including subnets, internet gateway, route tables, security groups, and VPC peering, to control access between private and public blocks.

Create a complete VPC setup with subnets, route tables, internet gateway, NAT gateways, and S3 endpoints using VPC and more; configure CIDR ranges, IPv6 options, and AZ-based subnets.

Define and analyze artificial intelligence, its benefits, trends, and industry applications, including visual perception, speech recognition, decision making, and language translation, with emphasis on daily life integration and upskilling.

Understand the ai lifecycle’s seven stages from problem identification to deployment and monitoring. Learn data collection, cleaning, model development, evaluation, and production integration.

Explore generative AI, a subcategory of AI that creates text, images, and music from large training data. Understand its applications in creative arts, healthcare, gaming, and future trends like realism.

Explore AWS bedrock with party rock, try building apps free without accounts or cards, and generate country-based menus and widgets using diverse models like Titan, llama, and stability AI.

Explore generative ai techniques and models—transformer, diffusion, gans, vaes—applied to chatgpt-like and multimodal tasks, from natural language and code generation to high-fidelity image creation and synthetic data.

Compare foundational models with large language models, noting that LLMs are text-focused subsets of broader foundation models trained on diverse data and capable of contextually aware outputs.

Explore how natural language processing enables machines to understand, interpret, and generate text or speech, enabling translation, sentiment analysis, and chatbot interactions.

Amazon Bedrock provides a fully managed, one-stop platform for pre-trained foundation models from top providers, with training and fine-tuning for your models and AWS integration for content generation and chatbots.

Enable model access in AWS bedrock by selecting enable all models or individual models, provide organization details and use-case descriptions, and manage cross-region references for inference models.

Explore how the text playground provides a single input, output interaction, contrasting with chat, and learn to run I21 model examples in a text-based workflow.

Explore the image playground to compare text-to-image generation with OpenAI ChatGPT foundation models, using stability AI and Amazon Titan. Compare speed and image quality, noting the default three-image output.

Explore how inference parameters shape text generation by adjusting temperature, top p, and top k to balance predictability and creativity in foundation models.

Learn how temperature governs creativity and predictability in ai outputs, with 0–1 values and a 0.5 median. See how randomness versus structure shapes storytelling with anthropic cloud and meta 70B.

Explore how top k configurations limit responses and steer generation by combining temperature and probability with top p and top k settings to yield reliable, shortlists of results.

Tune temperature, top P, and top K to balance randomness and accuracy for the most probable solution with parameters, showing how maximum and minimum settings yield creative versus predictable outputs.

Learn to tune AI model outputs with additional configurations like system prompts, max length, stop sequences, beyond temperature, top P, and top k, to control persona and cost.

Safeguards protect information by using guardrails and watermark detection to ensure safe, reliable, and ethical use of generative ai; guardrails block certain topics, while watermark detection flags ai-generated images.

Set up guardrails to block sensitive topics and enforce policy, configuring topics, block messages, kms encryption, harmful categories, pii, regex, grounding, and testing to ensure safe model responses.

Explore applying guardrails in the chat option, update and version guardrails as policies evolve, and manage working drafts and versions with model selections like Anthropic Sonnet.

Master prompt management with a builder tool to create, test, and adapt prompts using variables and topics, enable encryption, and compare models for AI industry use cases.

Explore the rag process to build knowledge bases by crawling data from sources, embedding it with models, indexing in a vector store, and enabling retrieval by an LLM.

Master crawling and indexing to understand how data is discovered, stored, and retrieved by search systems, using crawlers, spiders, and scheduling.

Convert text, images, and audio into embeddings with embedding models to generate vectors, then store them in a vector db for semantic search, retrieval, clustering, and classification.

Explore retrieval augmented generation (RAG), a machine learning approach that combines information retrieval with text generation to produce accurate, context-aware responses.

Learn how to create an IAM user, set admin access, and build a knowledge base using Bedrock, syncing data sources and embedding models to answer questions from provided data.

Discover how agents interpret natural language to automate tasks, retrieve data, generate content, and power customer interactions across APIs, databases, and AWS bedrock.

Learn to create and use agents as task-oriented assistants, choosing models, configuring manual or assisted workflows, and linking knowledge bases, actions, and guardrails to accomplish bookings and recommendations.

Explore how the assistant creates agents with the agent builder by crafting prompts and action groups for travel tasks, including flights, hotels, car rentals, and preparing and testing the agent.

Explore prompt flow as a drag-and-drop pipeline for designing and orchestrating prompts for generative AI, including conditional data paths, prompt chaining, testing, and reusability.

Bedrock Studio is a fully managed, drag-and-drop console for building and managing generative AI apps on AWS, with pre-trained models, fine-tuning options, and prompt flows.

Identify the right fm for your use case by aligning tasks like text generation and image recognition, and weighing accuracy, precision, recall, and cost.

Explore automatic, AWS managed, and bring your own team evaluations, then set the model, task type, and metrics, using built-in or S3 datasets.

Master prompt engineering by crafting clear instructions, context with background information and examples, and persona settings, then iteratively refining prompts to guide foundation models toward accurate, relevant responses.

Master prompt design by identifying the four elements: instruction, context, input data, and output indicators, and applying zero-shot, few-shot, and chain-of-thought techniques for accurate model responses.

Discover the zero-shot technique, where a language model outputs results from direct task instructions without examples. See a hands-on demonstration of prompting and classifying text.

Explore the few-shot prompting technique, guiding a language model with task-and-output examples to shape responses, demonstrated through sentiment classification using positive, negative, or neutral labels.

Explore the chain-of-thoughts technique, revealing intermediate reasoning steps as tasks break into steps, and apply zero-shot and few-shot prompting with think step by step in a hands-on demo.

Explore how prompts shape foundation model responses in bedrock by adjusting the input to influence outputs, using a llama three chat example to show step-by-step math reasoning and personality prompts.

Analyze prompt misuse, including prompt injection and prompt leaking, and explain safeguards to prevent data breaches, unethical outputs, and incorrect results in LM models.

Learn to mitigate prompt misuse by applying content filtering, prompt validation, bias mitigation, monitoring and logging, and access management to safeguard AI deployments.

Explore Amazon Business queue, an enterprise AI assistant that uses your data to answer questions, summarize content, generate documents, and surface actionable insights.

Explore the Amazon Q workflow architecture, from data ingestion via connectors and permissions to identity management, query filtering, and generating responses using data sources, built-in knowledge, or plugins.

Learn how the IAM organization centralizes account management, consolidates billing, and enables cross-account access with service control and tagging policies for secure, compliant, and efficient resource access.

Create and manage AWS organizations, invite or add accounts, enable consolidated billing, organize accounts into units, and apply service control and tag policies to govern EC2 usage.

Learn how IAM Identity Center centralizes user management and access across AWS accounts and applications, providing single sign-on, permission sets, and streamlined onboarding.

Enable IAM Identity Center and create an organization instance to centrally manage users and multi-account access. Configure identity sources, permission sets, and applications for the use case.

Enable the IAM Identity Center in a single AWS account to auto-create and manage an AWS organization from the Identity Center. Remove any existing organization, then enable in North Virginia.

Create and manage users in the IAM Identity Center to enable access to AWS resources, add users individually or in groups, assign emails and display names, and generate one-time passwords.

Enable new users in the IAM Identity Center by activating seven-day invitations, setting passwords, enrolling MFA with a built-in or authenticator app, and assigning permissions.

Explore the amazon cube business ui to create and manage internal apps, try the quick application, configure service roles and encryption, and apply tags.

Learn how to add users to Amazon Cube using the IAM Identity Center and import existing users to assign subscriptions, highlighting centralized user management.

Compare lite and pro subscription plans in Amazon queue, with lite offering Q&A from knowledge bases and single sign-on. Pro unlocks file uploads, apps, custom plugins, and QuickSight BI insights.

Explore retrieval in an LM workflow using Amazon queue with Kendra or native indexes, fetch data in real time, supply context to the LM, and deliver succinct, relevant responses.

Identify and configure data sources—from S3, PDFs, Confluence, and GitHub—to enable retrieval, indexing, and cross-source access using Amazon Cube and Kendra.

Upload files to create a data source for your application, then index and process them to generate outputs for your use case in the Amazon Q business app.

Explore guardrails to block words and topics, protect confidential data, and control file uploads, data sources, and topic-specific responses. customize guardrails by user groups to enforce policy.

Learn to block words in the Amazon Q app using admin controls and guardrails, preventing blocked keywords from user prompts while demonstrating with examples like driving a car.

Explore Amazon Q developer, an AI-powered copilot for AWS, offering code completion, inline suggestions, debugging, security checks, and chat for architecture and resource queries.

Explore how the Amazon Q developer in the AWS console enables chat, error diagnosis, support tickets, and architecture guidance with SDK/CLI documentation and chatbot integrations.

Discover how to use the AWS console with Amazon Q developer to query service information, view billing details, inspect EC2 resources, manage cross-region access, and explore SageMaker features.

Diagnose and resolve AWS console errors with Amazon Q, focusing on EC2 placement group issues and supported instance types, with guidance for EKS, Lambda, and S3.

Learn how to raise an AWS support ticket directly from the AWS console using Amazon Queue, including drafting a case, selecting service and category, attaching files, and choosing contact options.

Explore how Amazon Q developer generates Python code in the console for calculator use cases, covering plus, minus, multiply, divide, and division by zero, with banking scenarios.

Learn to code with the Amazon Q developer in lambda or directly through Q, using boto3 to access Bedrock and invoke models, including a calculator example.

Learn how AWS chatbot integrates Slack, MS Teams, and Chime to monitor AWS resources, receive alerts, and run CLI commands from chat, enabling incident management and automated workflows.

Learn to integrate Amazon Cube with Bedrock and Lambda, use Boto3 to list foundation models, create a Bedrock function, and test end-to-end model access.

Explore AWS Cairo, an AI-powered coding assistant that uses natural language to generate production-ready code from prototype to deployment, accelerating development and simplifying learning.

Explore the Cairo workflow and its four components—steering, specs, hooks, and MCP—to transform natural language prompts into structured specs, coordinated design, automated maintenance, and seamless external tool integration.

Install Cairo.dev on macOS and configure a vscode-like IDE, sign in with Google, GitHub, or AWS, and import settings from VS Code to start projects with extensions.

See how to write a simple calculator code in Python using the Amazon Q KyroDev chat model, with model selections, autopilot options, and free credits.

Learn how Cairo helps you understand code by cloning a GitHub repo and inspecting node.js projects, explaining files like app.js and index.css and guiding documentation generation and feature building.

Create and formalize a spec to structure project artifacts, including a Node.js project with folder layout and package.json scripts, and learn to install, start, run tests for learners.

Configure MCP with AWS Labs via Cairo CLI to load MCP servers and access AWS resources like Bedrock, Kendra, and Keyspaces from GitHub, enabling streamlined external integrations.

Discover how machine learning, a data-driven subset of artificial intelligence, trains models with supervised and unsupervised learning and applies inference on data to predict outcomes in healthcare, finance, and e-commerce.

Discover the four data types—structured, unstructured, semi-structured, and time series—and how schemas, timestamps, and formats like SQL, JSON, XML, and DynamoDB shape AI models.

Learn the machine learning development lifecycle, from identifying business goals and framing the problem to data collection, cleaning, labeling, feature engineering, training, deployment, and monitoring.

Explore supervised learning with labeled data, unsupervised learning that discovers patterns in unlabeled data, and reinforcement learning with reward-based learning and human feedback.

Supervised learning uses labeled data to train models that predict or classify inputs from predictors to targets, using training and testing phases to assess performance.

Explore supervised learning algorithms—from linear and logistic regression to decision trees, random forests, and k-nearest neighbors—plus gradient boosting with xgboost and lightgbm.

Explore unsupervised learning by analyzing unlabeled data to discover hidden patterns and group items by features, using clustering, dimensionality reduction, and anomaly detection (k-means, DBSCAN, PCA, SNE).

Learn how reinforcement learning uses reward-based trial and error to train an agent that acts in an environment, refines its policy, and applies to robotics, healthcare, and finance.

Learn how reinforcement learning with human feedback (rlhf) aligns foundation models to desired outcomes through a reward model. Feedback guides updates and fine-tuning to improve alignment and reduce bias.

Explore hyperparameters, the external configurations set before training, and how they control learning rate, batch size, epochs, dropout, and regularization (L1, L2) to prevent overfitting and improve performance.

Explore model evaluation metrics across classification, regression, and clustering, including precision, recall, F1 score, ROC AUC, MAE, MSE, and adjusted rand index, to assess performance and guide hyperparameter tuning.

Examine GenAI model performance metrics, from accuracy and cross-domain efficiency to conversion rate, revenue, and customer lifetime value, and explore ROC, BLEU, GLUE, HELM, MLU, and Big Bench evaluations.

Discover how MLOps blends machine learning and DevOps to automate end-to-end ML workflows through CI/CD, enabling scalable, reliable, reproducible production models and cross-team collaboration.

Explore SageMaker, AWS's end-to-end machine learning platform, covering data prep, labeling, training, tuning, deployment, and monitoring with no-code and notebook options, supporting multiple frameworks.

Learn to set up a SageMaker domain as a workspace, manage user access and resources, and use Studio to explore data prep, AutoML, pipelines, and endpoints.

Explore AWS SageMaker AI services across the ML lifecycle, including data prep with data wrangler and ground truth, building with Studio and Canvas, and monitoring with Clarify and Model Monitor.

Harness the data wrangler to import, clean, and transform data using canvas and Wrangler tab. Visualize insights with graphs, explore sample loan data, and build data flows without coding.

Discover SageMaker Jupyter in AWS studio, create and open Jupyter spaces, and run Python and PySpark notebooks for practical machine learning coding.

Explore how to use SageMaker AutoPilot in the canvas environment to build, validate, preview, and deploy pre-built models with automl, selecting data, target columns, and evaluating root mean square error.

Explore SageMaker Canvas, a no-code tool to build models directly from the studio, offering ready-to-use options for document queries, sentiment analysis, and text or image analysis.

Explore SageMaker JumpStart to deploy pre-built models from providers like hugging face, bedrock, and meta ai, using notebooks and jupyter lab for quick text generation workflows.

Discover SageMaker training models, including scalable model training, experiment tracking, and hyper ports for distributed training. Upload data to S3, start training jobs, and monitor metrics to optimize large-scale workloads.

train models using SageMaker experiments, tune hyperparameters and scalable clusters, explore MLflow and MLflow experiments, and manage S3 URIs and Studio workflows for end-to-end model training.

Deploy real-time machine learning models with SageMaker endpoints. Upload trained models to S3, configure and deploy the endpoint, and invoke it for predictions via API.

Explore SageMaker monitoring, including model monitoring, data drift detection, and baseline vs live predictions with CloudWatch alerts, plus Clarifai bias detection and Shapley additive explanations for explainability.

Master SageMaker Clarify hands-on for model evaluation, including automatic and human-based evaluation. Learn how to set up teams, provide instructions, and review performance metrics such as latency and throughput.

Explore SageMaker governance to ensure transparency, accountability, and compliance across the machine learning lifecycle, from model cards and centralized documentation to model registry and lineage tracking.

Explore SageMaker model cards in a hands-on use case, providing a centralized governance view with a dashboard that summarizes models, endpoints, and evaluation metrics to ensure transparency.

Master SageMaker inference for real-time and batch predictions by deploying trained models to endpoints, auto-scaling, cost-efficient hosting, and monitoring with CloudWatch and drift detection.

Explore Amazon Comprehend, a natural language processing service by AWS, that extracts insights from unstructured text with sentiment, entities, key phrases, language detection, and confidence scores.

Learn how AWS OpenSearch, a fully managed vector search and analytics service built on OpenSearch and Elasticsearch, enables fast search, analytics, visualization, and AI model integrations with SageMaker and Bedrock.

Explore how Amazon QuickSight Q enables NLP-based, real-time dashboards with instant insights, reducing manual queries and enabling cost-effective self-service analytics.

Explore amazon translate, a neural machine translation service by AWS, enabling real time translation of text across more than 75 languages with custom dictionaries and large-volume support.

Explore Amazon Transcribe, an AWS automatic speech recognition service delivering real-time and batch transcripts with time steps and speaker diarization, customizable vocabulary and language support.

Explore Amazon Lex, a managed conversation bot service using utterances and intents. It uses natural language understanding and integrates with Amazon Connect, AWS Lambda, ASR, and Bedrock or traditional models.

Explore Amazon Rekognition, a managed service that analyzes images and videos to detect objects, scenes, facial analysis and recognition, text, celebrities, and activities, with real-time insights and AWS integration.

Discover Amazon Forecast, a fully managed ML service in SageMaker Canvas that uses historical data to automate data preparation and model building, producing scalable time-series forecasts and dashboards.

Amazon Personalize delivers real-time recommendations using user events, item metadata, and user metadata. It integrates easily with apps, requires no ML expertise, and offers pay-as-you-go, recommendations for products and content.

Learn how Amazon Textract automatically extracts text, handwritten or printed, along with tables and form data, from scanned documents to reduce manual data entry.

Amazon Kendra is an intelligent retrieval service powered by machine learning. It enables high-accuracy indexing and information discovery across data sources with natural language processing and a unified search experience.

See how Amazon augmented AI blends machine learning with human review to improve accuracy, compliance, and fairness across industries.

Explore AWS Inferentia and Trainium, AI chips designed for deep learning inference and training that deliver low latency, high throughput, and cost efficiency on EC2 with the AWS Neuron SDK.

Apply responsible AI by upholding fairness, explainability, robustness, privacy, and transparency while ensuring ISO and HIPAA compliance, audits, and documentation to prevent harm and build trust.

Examine GenAI risks like hallucinations that generate false data and erode trust. Assess intellectual property issues, bias, toxicity, and privacy concerns, plus guardrails and GDPR compliance.

Understand the importance of model transparency and why regulators require it. Explore interpretability and explainability, compare linear models with neural networks, and study the trade-offs between transparency and performance.

Explore tools for model transparency, including open source models like Hugging Face transformers and AWS AI service cards, plus SageMaker model cards for documenting capabilities and limitations.

Explore data governance as a framework for data integrity, security, and usability, covering data curation, understanding, protection, profiling, cataloging, and lineage to ensure GDPR and HIPAA compliance.

Identify global AI compliance standards, including ISO 4201 and CSA, with region-specific rules. Learn AI risk management through RMF, governance, and the Algorithmic Accountability Act for transparency and bias monitoring.

Enable data governance with AWS Glue DataBrew by cleaning, normalizing, and transforming data without coding, using automated quality checks, pre-built transformations, masking sensitive data, and seamless S3 integration.

Set clear governance objectives for fairness, accountability, transparency, and compliance. Assemble a multidisciplinary team, monitor data quality, and ensure explainability and audits across the model life cycle.