[NEW] Ultimate AWS Certified AI Practitioner AIF-C01

Explore how the AWS cloud underpins AI workloads and how managed AI services like Amazon Bedrock operate, in the context of the AWS certified AI practitioner exam.

Learn how cloud computing delivers on-demand compute, memory, storage, databases, and networking, with IP addresses, routers, switches, and DNS, replacing traditional on-premises data centers.

Discover the on-demand, pay-as-you-go model that powers instant provisioning of compute, storage, and applications with AWS. Learn how cloud types—public, private, and hybrid—offer scalability, cost efficiency, and global accessibility.

Identify and compare the three cloud service models—IaaS, PaaS, and SaaS—and understand how AWS pricing for compute, storage, and networking enables cost savings.

Trace the evolution of the AWS cloud from its 2002 origins to public services like SQS, S3, and EC2, and map its global regions, availability zones, and edge locations.

Create an AWS account on a free plan, verify identity with a root user email and phone, set a compliant password, and access the console for hands-on learning.

Discover how the AWS console's new UI adopts rounded, blue buttons and a bright white layout while preserving the same usability as the old gray, square-button interface.

Explore the AWS console, select the nearest region to reduce latency, and use the services list or search bar to compare global versus regional availability.

Clarify the shared responsibility model, defining your security duties in the cloud versus AWS's. Prepare for exam questions about responsibilities.

Understand AWS AI services pricing, track charges, and learn to turn services off to avoid costs. Set up a course budget with email alerts to monitor spending and receive notifications.

Learn to set up a budget and alarms in the AWS billing console, view cost data and bills, and use zero spend and monthly budgets with alerts.

Explore generative AI and Amazon Bedrock, the main AWS service for generative AI, highlighting its role as a key exam topic and one of the fastest growing AWS services.

Discover what generative AI is, from foundation models and large language models to diffusion-based image generation, and understand non-deterministic outputs. Explore examples like GPT-4o and AWS Bedrock.

Explore Amazon Bedrock, a fully managed service for building generative AI apps with a unified API, foundation models from multiple providers, and data staying in your account.

Explore Amazon Bedrock by browsing the model catalog, testing models in chat and image playgrounds, and comparing Nova Micro with Claude Sonnet 4.5.

Explore how to choose a base foundation model on Amazon Bedrock by weighing performance, customization, context windows, and token limits across Amazon Titan, Llama-2, Claude, and Stability AI.

Compare Amazon Bedrock foundation models and other providers, evaluating capabilities, image support, and output formats; explore customization via supervised fine-tuning, reinforcement fine-tuning, and distillation for specific tasks.

Fine-tune a foundation model in Amazon Bedrock using S3 data; explore supervised and reinforcement fine-tuning, distillation, and cost considerations for tailored, efficient ai models.

Evaluate foundation models on Amazon Bedrock using automatic evaluation and benchmark datasets. Compare model outputs to benchmark answers with metrics such as rouge, bleu, bertscore, and perplexity to gauge quality.

Evaluate Bedrock models using automatic and human methods, including programmatic and model-as-judge approaches, for tasks like text generation and summarization, with toxicity, accuracy, robustness metrics, and S3 storage.

Explore how retrieval-augmented generation uses a knowledge base and vector embeddings to fetch up-to-date data and augment prompts for accurate answers within Amazon Bedrock.

Practice knowledge bases in RAG by uploading a document and chatting with it, using search results and a prompt template to answer questions with sources from your document.

Learn to set up a knowledge base and retrieval augmented generation on Amazon Bedrock, including creating an IAM user, configuring S3 sources, and building a vector store with OpenSearch Serverless.

Explore Gen AI concepts such as tokenization, token IDs, and context windows. Understand embeddings, vectors, vector databases, and nearest-neighbor search powering RAG and semantic relationships.

Explore guardrails in Amazon Bedrock to control interactions with foundation models, filter harmful content, block topics like food recipes, and monitor inputs to reduce hallucinations and protect privacy.

Explore guardrails to filter content with configurable filters, denied topics such as recipes, profanity, and PII removal, and contextual grounding to reduce hallucination, demonstrated in a hands-on test.

Explore Amazon Bedrock agents that think and act on multi-step tasks, integrate with APIs and Lambda, access knowledge bases, and use chain-of-thought reasoning to manage infrastructure and orders.

Explore how Amazon Bedrock integrates with CloudWatch. Log all model invocations to CloudWatch Logs or S3, including inputs, outputs, and embeddings, with real-time insights and guardrail metrics.

Integrate Amazon Bedrock with CloudWatch logs to enable model invocation logging and route data to CloudWatch or S3, then monitor latency, tokens, and model invocations via CloudWatch metrics.

Explore Amazon Bedrock pricing: on-demand, batch mode discounts up to 50%, and provision throughput; major costs come from input/output tokens and vector databases for RAG, while prompt engineering stays cheap.

Demonstrate an end-to-end Amazon Bedrock use case by implementing API calls to power the AI Stylist demo, using knowledge bases, AI agents, and generated outfits.

Explore Amazon Nova, an AI model family on bedrock, featuring Nova Premier to Nova Micro, Canvas for images, Reel for video, Sonic for speech, and Nova 2 Omni for multimodal tasks.

Master prompt engineering to excel on the exam and apply versatile skills to any LLM, including ChatGPT and cloud platforms, gaining a head start in the AI race.

Explore prompt engineering by designing prompts with instructions, context, input data, and output indicators, enhanced by negative prompting to guide foundation models toward clear, concise AWS summaries.

Practice building effective prompts using instructions, context, and input data to generate detailed itineraries with times, locations, and dining, while applying negative prompting to refine results.

Unpack prompt performance optimization for LLMs by tuning system prompts, temperature, top P, top K, length, and stop sequences, and note latency depends on model size, type, and context window.

Practice configuring models like Anthropic's Claude 3 Sonnet to influence creativity by adjusting temperature, top P, and top K, comparing low and high settings in a robot cooking story.

Explore zero shot prompting, few shots prompting, one shot prompting, chain of thought prompting, and retrieval-augmented generation (RAG) to improve prompts with external data sources.

Explore how prompt templates standardize inputs and outputs, enable few-shot prompting, and protect against injection attacks, using placeholders and Parity Rock demonstrations.

Discover Amazon Q and its potential to unlock new use cases with your internal data. Observe how it reshapes interaction with the AWS cloud as it grows more powerful.

Learn how Amazon Q Business leverages gen-ai trained on your company data with data connectors and plugins to search documents, automate tasks, and enforce security via IAM Identity Center.

Practice building a generative application with Amazon Q Business, using anonymous access, adding an S3 data source, creating an index, and syncing a knowledge base with guardrails.

Amazon Q Apps enables gen AI-powered apps without coding via natural language, using Amazon Q Apps Creator, built from your company data with document uploads and plugins.

Discover Amazon Q Developer, an AWS assistant that answers questions about documentation and resources and acts as a code companion. Lists Lambda functions, suggests CLI commands, and analyzes bills.

Explore Amazon Q and Amazon Q Developer, the AI coding assistant for AWS, with cross-region data access, S3 bucket management via UI and CLI, CloudShell, and bill insights.

Discover how Amazon Q enables natural language queries across AWS services, from QuickSight dashboards to EC2 instance recommendations and Glue ETL support via the AWS Chatbot.

PartyRock offers a playground without an AWS account, not a real AWS service, powered by Amazon Bedrock to build gen AI apps with widgets and prompts, restaurant recommendations, recipe ideas.

Explore the behind-the-scenes ideas of AI, ML, DL, and generative AI in a theory-oriented section that goes beyond concrete AWS services, preparing you for the exam.

Learn how AI encompasses ML, deep learning, and GenAI, powered by transformer models and diffusion models, with data, model, and application layers guiding use cases like computer vision and NLP.

Learn a quick executive summary of key machine-learning terms like GPT, BERT, RNN, ResNet, SVM, WaveNet, GAN, and XGBoost, focusing on their intents for exam preparation.

Learn how training data drives machine learning and why clean data matters. Explore labeled versus unlabeled data, supervised and unsupervised learning, and structured and unstructured data formats.

Explore supervised learning by building mappings from labeled data to predict outputs, covering regression and classification, and learn how training, validation, and test sets enable model evaluation and feature engineering.

Explore unsupervised learning on unlabeled data, where algorithms group inputs into clusters and reveal patterns, with techniques like clustering, association rule learning, anomaly detection, fraud detection, and semi-supervised learning.

Learn self-supervised learning with unlabeled data where a model creates its own pseudo-labels through pre-text tasks. Discover how this approach builds internal representations for downstream tasks like summarization.

Describe how reinforcement learning trains an agent to maximize cumulative reward by navigating an environment like a maze, using states, actions, rewards, and a policy learned through simulations.

Explore reinforcement learning from human feedback (rlhf) to align models with human goals by using human ratings to train a reward model and optimize language models.

Explore model fit, bias, and variance and how they affect training and unseen data. Identify overfitting, underfitting, and balanced models, and reduce bias and variance with feature choices.

Explore model evaluation using a confusion matrix for classification, with precision, recall, f1, and auc-roc, then assess regression with mae, mape, rmse, and r-squared.

Explore real-time, batch, and edge inferencing to understand how models make predictions on new data. Assess trade-offs between edge devices, local inference, and remote servers to optimize latency and accuracy.

Identify business problems and translate them into machine learning goals; prepare data, engineer features, train and tune models, evaluate, deploy, monitor, and retrain with data or feature augmentation as needed.

Explore hyperparameters that define model structure and learning: learning rate, batch size, epochs, and regularization. Use grid or random search and SageMaker AMT to improve accuracy, generalization, and reduce overfitting.

Explore when ML (supervised, unsupervised, or reinforcement learning) is unsuitable for deterministic problems and prefer exact code solutions for the exact answer, even as language models improve.

Explore AWS specialized AI services for image recognition, text translation, and speech generation to prepare for the exam.

Explore AWS AI managed services—from Bedrock GenAI and SageMaker to Comprehend and Lex—and see how token-based pricing lets you pay only for what you use in ML.

Discover Amazon Comprehend, a fully managed serverless nlp service that uses machine learning to extract key phrases, named entity recognition, sentiment, and topics from text, with custom classification.

Discover Amazon Comprehend, a natural language processing tool that extracts entities, PII, dates, and sentiment from text, enables syntax analysis, and supports custom classification with training data.

Explore Amazon Translate, a natural and accurate language translation service that localizes content for international users and efficiently handles large text volumes.

Explore Amazon Translate's neural translation service, translate texts and documents between languages, run batch jobs, and customize results with terminology and parallel data.

Discover Amazon Transcribe's automatic speech recognition that converts audio to text, with PII redaction, multilingual language identification, custom vocabularies, domain-specific language models, and toxicity detection for searchable transcripts and metadata.

Explore Amazon Transcribe to generate transcripts from audio, enable PII redaction, and stream in multiple languages with automatic language identification, including English and French.

Learn how Amazon Polly turns text into lifelike speech using deep learning, with lexicons, SSML, neural, standard, long-form, and generative voices, and speech marks for lip-sync and highlighting.

Experiment with Amazon Polly's text-to-speech engines: generative, neural, and standard, and test SSML for breaks, pauses, pronunciation, and different output formats.

Explore amazon rekognition’s image and video analysis, including face detection, face search and verification, text and label detection, celebrity recognition, and custom labels for logos, plus automated content moderation.

Explore Amazon Rekognition's capabilities—label detection with object and custom labels, image properties, moderation, facial analysis and comparison, text in image, celebrity recognition, and PPE detection.

Build chatbots with Amazon Lex that interface by voice or text, infer intents. Use slots to collect data and invoke Lambda functions for hotel bookings, with multi-language support.

Explore Amazon Lex to build chatbots and conversational AI, using traditional and generative approaches, with intents, utterances, slots, and optional Lambda integrations in a visual builder.

Explore Amazon Personalize, a fully machine learning service for real-time personalized recommendations built with ready-to-use recipes. Use cases span retail, media, and entertainment, with data from S3 and real-time APIs.

Explore how Amazon Textract extracts text, handwriting, and data from scanned documents using AI, reading PDFs and images, including forms and tables for invoices, medical records, and tax forms.

Explore Amazon Textract hands-on to analyze documents like pay stubs, extract raw text, identify layout elements, and pull key-value pairs, tables, and ID fields.

Utilize Amazon Kendra, a fully managed document search service powered by machine learning that extracts answers from documents and enables natural language search with incremental learning and fine-tuning.

Explore Amazon Mechanical Turk, a crowdsourcing marketplace that assigns simple human tasks to a distributed workforce, enabling image labeling and data collection with integration to Amazon A2I.

Discover Amazon Augmented AI (A2I): high-confidence predictions return instantly, while low-confidence cases go to human review, with risk-weighted scores stored in Amazon History and fed back to improve models.

Explore Amazon Augmented AI in the SageMaker console to create human review workflows for Rekognition and Textract outputs, including image moderation and low confidence triggers.

Explore amazon transcribe medical and comprehend medical to convert physician dictations into structured, phi-aware insights, enabling real-time or batch transcription and medical data analysis.

Demonstrates Amazon Comprehend Medical and Transcribe Medical in real time to extract entities, relationships, and meds from doctor notes and medical discussions.

Use AWS HealthScribe to automatically generate clinical notes from patient conversations, with speaker role identification and medical term extraction. Streamline documentation and patient visit recaps with HIPAA-eligible transcripts.

Explore Amazon EC2, virtual servers in the cloud, with GPU options, Trainium and Inferentia chips, and scalable features like ASG and ELB for AI workloads.

Explore AWS EC2 instance types for machine learning with get advice for training and inference. Learn about trn1, g5g, c7gn, inf2, and inf2.48xlarge, NVIDIA T4G GPUs, and their hourly pricing.

Explore Amazon SageMaker at a high level for the AWS certified AI practitioner exam, highlighting key features relevant to data scientists and data engineers and exam-ready concepts.

Explore Amazon SageMaker, a fully managed machine learning service that builds, trains, tunes, and deploys models, with real-time, serverless, asynchronous, and batch inference and SageMaker Studio for end-to-end development.

Explore SageMaker AI Studio to build, train, and deploy models at scale, using JumpStart base models, datasets, evaluators, and pipelines within a single domain user interface.

Discover how SageMaker Data Wrangler and SageMaker Feature Store streamline data preparation, transformation, feature engineering, and exporting data flows for machine learning in SageMaker Studio.

Assess foundation models with SageMaker Clarify and Ground Truth, comparing models on tasks, bias, explainability, and human preferences through RLHF and human-in-the-loop evaluation.

Leverage SageMaker governance tools to manage deployed models with model cards, model dashboard, and role manager, and monitor quality via model registry, pipelines, and drift checks.

Explore SageMaker JumpStart’s machine learning hub, deploying pre-trained models from providers like Hugging Face and Meta, with no-code canvas and MLFlow integration for managing the machine learning lifecycle.

Explore Amazon SageMaker, an end-to-end machine learning service featuring automatic tuning, flexible deployment, SageMaker Studio, Data Wrangler, Feature Store, Clarify, Ground Truth, Model Cards, and Pipelines.

Explore SageMaker extra features, including network isolation mode to restrict outbound access and protect training data, and the DeepAR forecasting algorithm that uses an RNN to forecast time series data.