AWS Certified AI Practitioner AIF-C01 Masterclass

Explore when to apply ai and ml to boost business efficiency with dynamic pricing, predictive analytics, risk management, and when not to use. Learn factors like cost, resources, and interpretability.

Explore supervised and unsupervised learning types, including classification and regression, clustering with k-means, and anomaly detection using statistical methods like z-scores and isolation forest.

Explore AWS managed ai and ml services for text and document analysis, including Amazon Comprehend and Textract, plus Transcribe, Lex, Polly, and Rekognition for speech, chatbots, text-to-speech, and image/video analysis.

Explore aws managed ai/ml services for customer experience, including amazon personalize for real-time recommendations, kendra for intelligent search, translate for neural translation, and forecast and fraud detector.

Discover real world ai applications powered by AWS managed services, including Rekognition visual search, Comprehend sentiment analysis, Transcribe subtitles, Personalize recommendations, Fraud Detector, and Forecast for demand.

Explore AWS SageMaker, a fully managed ml service that streamlines data preparation, training, deployment, real-time inference, monitoring, and end-to-end ml ops across the lifecycle.

Discover data collection, pre-processing, feature engineering, and data processing for machine learning, and learn how AWS Glue and SageMaker automate ETL, data cataloging, and data preparation.

Explore Amazon SageMaker options including Ground Truth for automated and human labeling, Canvas for visual feature engineering with 300+ transformations, and SageMaker Feature Store for centralized feature management.

Train, tune, and evaluate machine learning models with SageMaker by running training jobs, comparing algorithms, and hyperparameter tuning to optimize metrics like AUC.

Learn to select batch or real time inference, expose predictions via rest api interfaces with api gateway and lambda, and manage containerized model artifacts through SageMaker endpoints.

Monitor machine learning models, detect data drift and concept drift, and implement MLOps with SageMaker pipelines, model registry, and deployment monitoring for quality data.

Explore how confusion matrices evaluate model performance, detailing true and false positives and negatives, and measure accuracy, precision, recall, F1 score, specificity, and false positive rate.

Explore ROC curves and AUC to evaluate thresholds in model performance, then compare regression errors via MSE, RMSE, and MAE alongside ROI considerations.

Generative AI processes vast internet data using a transformer model to understand information and generate new content. It combines understanding and creation to produce text, images, and code.

Compare traditional AI and generative AI, noting content creation, non-deterministic responses, and randomness control. Recognize training costs and vendor APIs, while traditional AI handles reasoning and generative AI handles creativity.

Understand how GPT uses a transformer to process tokens, pre-train on large data, and generate text for translation, summarization, and Q&A, with flexible data types like images and video data.

Trace the generative AI timeline from Google's attention is all you need transformer paper to GPT-3 and ChatGPT. Explore AWS Bedrock's serverless foundation models like Titan and Stable Diffusion.

Define the scope and use case of a generative AI project, choose a model, apply prompt engineering and in-context learning, then fine-tune, evaluate, deploy, and monitor for performance and drift.

Explore tokens in generative AI and how the OpenAI tokenizer breaks text into tokens, affecting input and output costs across GPT-4, GPT-3.5, and legacy models.

Explore how temperature controls randomness in generative ai, from low, predictable outputs to high, creative ones, and how tokens—via tokenization and vocabulary mappings—drive cost and performance.

Embeddings transform words, items, and concepts into unique numerical codes that preserve similarity and relationships. They enable better predictions, powering search, recommendations, and language understanding.

Master vector database concepts for embeddings used in generative AI, enabling fast, scalable storage and retrieval for semantic search, text and image search, product recommendations, and anomaly detection.

Generative AI model training uses fine tuning to specialize in a task, retrieval augmented generation to pull external knowledge, and few-shot learning to generalize from small examples.

Foundation models deliver versatile, cross-domain capabilities by pre-training on massive self-supervised data and fine-tuning with limited labeled data, enabling prompting, few-shot, and zero-shot learning.

Differentiate foundation models, generative AI, and large language models. Explain how foundation models act as base structures, how generative AI creates data, and how LLMs focus on language tasks.

Explore multi-modal AI, where text, voice, image, or video inputs feed a transformer to output text, images, or videos; unimodal LLMs handle only text.

Explore the transformer architecture and its self-attention mechanism that powers generative AI and long-term dependencies. Learn how positional encoding, query-key-value vectors, and multi-head attention drive next-token predictions.

Explore generative AI use cases such as text summarization, information extraction, translation, and developer assistance with code generation. Examine limitations like context retention, reliance on training data, and potential hallucinations.

Explore AI agents that autonomously reason, plan, and execute in dynamic environments, using tools, memory, and context to achieve goals with minimal human input.

Explore the AWS generative AI stack across hardware, model, and application layers, including foundation models infrastructure, and tools like Bedrock and SageMaker for Rag applications, content and code generation.

Explore AWS generative AI services, including Bedrock and SageMaker, to build applications, select foundation models, and leverage data foundation options like training from scratch, fine tuning, and rag-based in-context learning.

Discover the advantages of AWS generative AI services, including accessible, scalable, and cost-efficient models via pre-built options like SageMaker Jumpstart, with security, CAF AI guidance, high availability, and global infrastructure.

Explore five Amazon Q services: Amazon Q for business, QuickSight, Q developer, Amazon Connect, and Amazon Queue—enabling internal data training, AI-generated dashboards, and generative AI features across the suite.

Explore AWS SageMaker Jumpstart to quickly deploy and evaluate open-source and partner generative AI models, from Llama to Meta, via ready-to-deploy endpoints and JupyterLab notebooks.

Explore Amazon Bedrock, a fully managed service that provides foundation models via a single API with serverless deployment, private customization, and seamless AWS integration.

Explore Amazon bedrock UI options, including providers, base and custom models, knowledge bases for retrieval augmented generation, guardrails, agents, and prompt flows, with playground access.

Discover how temperature, top p, and top k shape model output and determinism, and see how combining these parameters influences the next token in practical examples.

Explore bedrock playground’s chat, text, and image generation, load and view 40+ examples with Python code and AWS CLI details, and learn model access and pricing.

Explore guardrails in AWS Bedrock to prevent prompt attacks and block content, and implement filters for harmful content, profanity, and contextual grounding.

Explore retrieval augmented generation by merging generative models with a vector database to deliver context-rich, accurate responses sourced from knowledge bases via offline embedding workflows.

Explore knowledge bases in Amazon Bedrock to implement RAG, configure data sources and embeddings, and test with a Paris Olympics 2024 prompt.

Explore the architecture and design flow of a typical generative ai application built on Amazon Bedrock. Learn how api gateway, lambda, boto3, and s3 handle prompts, foundation models, and responses.

Explore practical criteria for selecting pre-trained models, including cost, latency, modality, customization, size, input/output limits, transfer learning, privacy, security, bias, and ethical concerns, hardware constraints, and future replaceability.

Explore prompt engineering to shape generative AI responses via task instructions, context, and input text. Identify prompt types like chain of thought, zero-shot, and few-shot; quality improves coherence.

Explore zero-shot, one-shot, and few-shot prompting to train models, using sentiment classification examples to show how multiple prompts guide responses.

Explore prompt engineering basics, define versus explain prompts, tailor explanations to personas from five-year-olds to data scientists and software engineers, and generate versatile output formats for large language models.

Explore official promotion emails, draft responses using ChatGPT, build learning plans, and use reverse prompts to generate interior design prompts.

Explore seven popular prompt engineering techniques, including esyo, tag, tf, bab, care, rice, and chain of thought, with practical examples and hands-on ChatGPT demonstrations.

Master popular prompt engineering techniques—bab, care, rise, and chain of thought—through practical prompts for async file uploads via a message queue, reddis caching, ci/cd, and debugging.

Learn to construct a ten-component AWS prompt, covering task context, tone, background data, and task rules. Explore applying examples, conversation history, think step by step, and output formatting on bedrock.

Explore prompt engineering risks and limitations, including prompt injection attacks, jailbreaking and bypassing guardrails, hijacking prompts, and poisoning inputs, with guardrails to prevent leakage of confidential information.

Learn about foundation models training: pre-training on vast datasets to gain language and general reasoning, fine-tuning for domain-specific tasks, and continuous pre-training with periodic updates.