AWS Certified AI Practitioner AIF-C01 [Beginner to Advanced]

Machine learning, a subset of artificial intelligence, trains algorithms on historical data to build models that generate insights, covering supervised, unsupervised, and reinforcement learning with labeled data and clustering.

Explore the artificial intelligence landscape with a mental map, covering machine learning, supervised and reinforcement learning, neural networks, deep learning, generative ai, computer vision, nlp, and robotics.

Explore core computer vision capabilities—image classification, object detection, image segmentation, and facial recognition—and their real-world use cases in retail, healthcare image analysis, smart factories, and self-driving cars.

Explore Amazon Rekognition, a cloud computer vision service for content moderation, identity verification, and streaming video analysis, featuring label detection, image properties, facial analysis, and real-time tagging.

Explore natural language processing, a subset of artificial intelligence using machine learning to enable computers to understand human language, including chatbots, virtual assistants, translation, and sentiment analysis.

Explore how Amazon Lex builds conversational AI chatbots for voice and text, covering intents, utterances, slots, confirmation, and fulfillment, and integrating with Lambda to power hotel, flight, and enterprise bots.

Explore generative AI, a branch of artificial intelligence using foundation models to generate text, images, and code from trained data, with examples like ChatGPT and Llama on Amazon Bedrock.

Explore how prompts are tokenized and processed within a context window to generate the next word. Understand tokens, parameters, and temperature, plus open-source and proprietary foundation models.

Discover four broad streams of generative AI use cases. Explore text generation, summarization, and chatbots; image and video creation; code generation and upgrades; and embeddings for search.

Explore foundation models, including large language models, diffusion and multimodal systems, with examples from Amazon Bedrock, Stability AI, and Claude, and learn how text, image, and healthcare use cases unfold.

Learn how words and documents become vectors through embeddings, using Amazon Titan embedding model to create 1536-dimensional vectors, and perform semantic vector search with cosine similarity and ANN methods.

Build a hands-on AWS demo by creating an AWS Lambda function that sends prompts to Amazon Bedrock Titan embed text version 2.0, producing 1024-dimensional vector embeddings for a vector store.

Discover how Amazon Q developer functions as a generative AI powered assistant for software development, aiding coding and testing across Java, Python, .NET, and TypeScript, enabling upgrades and security scanning.

Explore Amazon Bedrock, a fully managed serverless generative AI service that provides API access to foundation models from AWS and partners, enabling prompt-driven requests with configurable temperature and token limits.

Explore the AWS Bedrock console, review base and custom foundation models, and learn to fine-tune, import via SageMaker, and deploy with provisioned throughput and guardrails.

Explore party rock, a bedrock-powered generative ai playground that lets you build apps with no coding, sign in with a social id, and experiment with prompts and foundation models.

Learn how Amazon SageMaker Jumpstart, a fully managed generative AI service within SageMaker, provides foundation models, computer vision models, and NLP models, enabling deployment, fine-tuning, evaluation, and scalable endpoints.

Explore AWS Trainium for training and Inferentia for deployment in a customer VPC, optimizing latency, throughput, and cost while building and fine tuning your own foundation models.

Launch an EC2 instance using Trainium and Inferentia for training hugging face transformers. The demo covers selecting the hugging face neuron deep learning AMI, VPCs, and security settings.

Explore how Amazon Bedrock agents act as intelligent orchestrators, routing user queries to data sources such as DynamoDB and RAC-based solutions, guided by foundation models and open API spec.

Build a retail banking agent on AWS Bedrock using Claude 3 Sonnet, illustrating pre-processing, orchestration, and post-processing with OpenAPI schemas, Lambda, DynamoDB, and a rag knowledge base.

Identify and prioritize generative AI use cases across an enterprise, from movie poster design with Stable Diffusion to multi-modal text summarization for technicians, using phase-wise workflows.

Evaluate foundation models for text summarization by comparing modality, size, latency, context window, pricing, training data, open-source versus proprietary options, fine-tuning, multilingual support, and response quality.

Assess machine learning and foundation model performance with accuracy, precision, recall, F1, ROC AUC; explore Rouge, Bleu, Bert Score, and Amazon Bedrock model evaluation service bring-your-own-team or AWS Manage Work.

Explore evaluating binary classification models with confusion matrices and metrics like accuracy, recall, precision, and F1 score, illustrated through Gmail spam and diabetes screening examples.

Explore ROC and AUC as model evaluation metrics, learn how to plot true positive rate against false positive rate across thresholds, and interpret area under the curve.

Explore foundation model evaluation metrics: rouge for automatic text summarization, bleu for machine translation, and BERTScore for semantic, embedding-based text generation assessment.

Learn to evaluate foundation models with AWS Bedrock's model evaluation service, using automatic, human, or AWS managed teams, with gigaword or exome datasets and metrics like accuracy, toxicity, and robustness.

Review high-level regression evaluation metrics, including mean absolute error, mean squared error, root mean squared error, and r-squared, with quick example explanations and their interpretation for exam readiness.

Master prompt engineering techniques from zero-shot, one-shot, and few-shot prompting with templates to improve foundation model outputs for e-commerce product descriptions.

Demonstrate hands-on fine tuning of foundation models, detailing model parameters, hyperparameters, datasets, batch size, and epochs. Create a fine-tune job with training data on S3, then evaluate using validation data.

Explore train and tune phase of the machine learning development lifecycle on AWS, from data processing and feature selection to training, evaluation, and hyperparameter tuning with SageMaker endpoints and containers.

Deploy and monitor phase uses the model registry, feature store, and container deployment to create an inference endpoint and monitor explainability, drift, and retraining triggers.

Discover how ML ops unifies ML development and operations to accelerate model delivery through people, processes, and technology, using an AWS end-to-end workflow.

Explore amazon sageMaker data Wrangler: import data from S3, Athena, and 50+ sources; assess data quality with exploratory data analysis; transform and export to S3 or SageMaker feature store.

Explore Amazon SageMaker clarify, emphasizing bias detection, explainability, and model evaluation for large language models. Discover how SageMaker Studio and notebooks streamline data prep, training, and deployment.

Explore how SageMaker model monitor tracks data quality and model performance in production, with drift alerts and retraining triggers. Review tools—model cards, model dashboard, and registry for versioning and approvals.

Automate machine learning workflows at scale with SageMaker pipelines for data prep, training, tuning, evaluation, and model registration, and use SageMaker Ground Truth with human feedback for labeling.

Explore how Amazon Polly converts text to lifelike speech using deep learning, enabling e-learning notes, content feeds, and call center applications, with integration to S3, AWS Lambda, and CloudFront.

Explore Amazon Transcribe, an AWS natural language processing service that converts speech to text, enabling real-time and post-call analytics, medical transcription, and toxic content detection.

Explore hyperparameters and tuning techniques for neural networks, including learning rate, batch size, and epochs, and compare grid search, random search, Bayesian optimization, and AWS SageMaker automatic model tuning.

Explore bias and variance, and how underfitting and overfitting affect model predictions. Learn to reduce bias with a more complex model and increase features; cut variance with feature selection.

Define responsible AI through eight guiding principles—veracity and robustness, safety, privacy and security, explainability, fairness, transparency, controllability, and governance—while exploring real-world challenges like hallucinations and toxic content.

Learn to apply responsible AI using AWS tools across data preparation to deployment, evaluating foundation models on Amazon Bedrock for toxicity, safety, and robustness, with guardrails, SageMaker Clarify, and governance.

Configure bedrock guardrails to block toxic topics, keywords, and PII with deny topics and custom user messages, and test these safeguards with console checks and watermark detection.

Explore how Amazon SageMaker Clarify detects data and model bias and provides explainability, showing pre-training and post-training bias, and feature influence, with model monitor tracking data quality in production.

Enable ML governance with Amazon SageMaker by using preconfigured role manager personas, model cards, and a unified model dashboard to support auditable, responsible AI across models.

Explore AWS identity and access management (IAM) for authenticating users, authorizing access to AWS services like S3 and EC2, and using policies, roles, and groups to manage permissions.

Learn hands-on how IAM roles, permissions, and policies authorize a Lambda function to access Amazon Bedrock, using a demo to attach the Amazon Bedrock full access policy and test access.

Learn to use AWS Artifact for on-demand compliance documents and AWS Audit Manager to map AWS usage to industry standards, collect evidence from CloudTrail and Security Hub, and automate audits.

Explore governance and regulation compliance with AWS config, CloudTrail, and CloudWatch. Learn how config monitors resource changes, CloudTrail tracks user activity, and CloudWatch analyzes metrics and logs.

Explore the generative AI security scoping matrix, outlining five scopes—from customer apps to self-trained models—and show how ownership and risk evolve across Bedrock and SageMaker.

Evaluate summarization quality by unigram overlap and rouge-based metrics, and learn why a maximum context window is crucial for summarizing long documents.

Walks through practice test questions on Amazon Bedrock CloudWatch logging, few-shot prompting, Bedrock pricing, AWS Audit Manager, and QuickSight Q for natural language bi dashboards.