AWS Certified AI Practitioner (AIF-C01)

Section 1: Lab: The AIF-C01 Exam and Preparing Your AWS Account

• Exam structure, question types, passing standard and the weighting of all five domains
• The AWS Free Plan, new-customer credits and which services need a Paid Plan
• Set up an IAM user with least privilege and choose a Region that offers Amazon Bedrock
• Lab: set an AWS Budgets alert before starting any other lab

Section 2: Core AI Concepts and Terminology

• How AI, ML, deep learning, generative AI and agentic AI differ
• Data types: labelled, unlabelled, tabular, time-series, images and text
• Supervised, unsupervised and reinforcement learning with examples
• Batch, real-time, asynchronous and serverless inference

Section 3: Lab: Matching AI to Business Problems

• Where AI adds value and when it should not be used
• Regression, classification, clustering, and when a foundation model beats traditional ML
• Managed AI services: Comprehend, Translate, Transcribe, Polly, Lex, Rekognition and Textract
• Lab: try managed AI services in the console with text, audio and documents

Section 4: The AI/ML Lifecycle and Amazon SageMaker AI

• Stages of an AI/ML pipeline from data collection to deployment and monitoring
• Managed APIs versus self-hosted models, and the roles of SageMaker AI and JumpStart
• MLOps concepts: repeatable experiments, monitoring and retraining when data changes
• Workshop: read a confusion matrix and calculate accuracy, precision, recall and F1

Section 5: Lab: Generative AI Fundamentals

• Tokens, chunking, embeddings, vectors, transformer-based LLMs and diffusion models
• Limits of generative AI: hallucination, nondeterminism and interpretability
• Token-based pricing and the role of context engineering
• Lab: compare answers from several models in the Amazon Bedrock playground and check token counts

Section 6: Agentic AI and Generative AI Services on AWS

• AI agent concepts: tools, memory, multi-agent patterns and the Model Context Protocol (MCP)
• Amazon Bedrock, Bedrock AgentCore, Strands Agents, Amazon Q, Amazon Quick and Kiro
• Cost factors: on-demand, provisioned throughput, Region and custom models
• Workshop: match the organisation's use cases to suitable AWS services

Section 7: Lab: Application Design and Prompt Engineering

• Model selection criteria: cost, modality, latency, language and the effect of temperature
• Zero-shot, few-shot, chain-of-thought, prompt templates and Bedrock Prompt Management
• Risks: prompt injection, jailbreaking, poisoning and data exposure
• Lab: tune prompts and inference parameters for a document summary task

Section 8: Lab: RAG with Amazon Bedrock Knowledge Bases

• What RAG is, how it reduces hallucination and where it fits in business
• Vector stores on AWS: OpenSearch Service, Aurora, RDS for PostgreSQL, Neptune and S3 Vectors
• Cost trade-offs of pre-training, fine-tuning, in-context learning, RAG and distillation
• Lab: build a knowledge base from documents in S3, ask questions with citations, then clean up

Section 9: Fine-Tuning and Model Evaluation

• Pre-training, fine-tuning, continued pre-training, instruction tuning and RLHF
• Preparing fine-tuning data: quality, size, labelling and governance
• Metrics: ROUGE, BLEU, BERTScore, LLM-as-a-judge and human evaluation
• Workshop: choose how to evaluate a RAG application or agent against business goals

Section 10: Lab: Responsible AI and Amazon Bedrock Guardrails

• Responsible AI principles: fairness, inclusivity, robustness, safety and veracity
• Risks around intellectual property, bias and trust in model output
• Transparency and explainability with SageMaker Model Cards and SageMaker Clarify
• Lab: create a guardrail that blocks denied topics and harmful content and masks personal data

Section 11: Security, Compliance and Governance

• The shared responsibility model, IAM, AWS KMS, Macie and PrivateLink for AI workloads
• AgentCore Identity and Policy in AgentCore for controlling what agents can do
• Auditing with CloudTrail, Config, Inspector, Artifact and Trusted Advisor
• Data governance: lineage, retention, logging and grounding to reduce hallucination

Section 12: Workshop: Review and Exam Planning

• Key content summarised by the weighting of each domain
• Practise scenario-style questions and how to rule out wrong options
• Free AWS revision resources and a study plan for after the course
• Check and delete every resource in your AWS account before the course ends