AWS Certified AI Practitioner Exam
65-question written exam with multiple-choice, multiple-response, ordering, and matching question types.
- Type
- Written
- Delivery
- Both
- Duration
- 90 min
- Questions
- 65
Passing score: 700 Scaled score on a 100 to 1,000 scale
Exam sections
Domain 1: Fundamentals of AI and ML
This domain covers fundamental AI/ML concepts including deep learning, NLP, and computer vision. Candidates must describe the ML development lifecycle, from data preprocessing and feature engineering to model training and monitoring using AWS services like SageMaker and specialized AI tools.
Question notes
Multiple choice, multiple response, ordering, and matching questions. No penalty for guessing.
Preparation tips
Review core AI/ML terminology and study the stages of an end-to-end SageMaker pipeline, including data preparation and model evaluation metrics like accuracy and F1 score.
Domain 2: Fundamentals of GenAI
Focuses on generative AI concepts such as tokens, embeddings, and transformer-based LLMs. It covers the foundation model lifecycle—selection, pre-training, and fine-tuning—and identifies AWS infrastructure for GenAI applications, including Amazon Bedrock and Amazon Q.
Question notes
Multiple choice, multiple response, ordering, and matching questions. No penalty for guessing.
Preparation tips
Practice prompt engineering and study the differences between foundation models; understand the business advantages and limitations like hallucinations and nondeterminism.
Domain 3: Applications of Foundation Models
Addresses the practical application of foundation models, including RAG and agentic AI. It covers prompt engineering techniques like chain-of-thought, fine-tuning methods such as RLHF, and evaluation metrics including ROUGE and BLEU to assess model performance against business objectives.
Question notes
Multiple choice, multiple response, ordering, and matching questions. No penalty for guessing.
Preparation tips
Build a simple RAG pipeline using Amazon Bedrock and experiment with various prompt engineering constructs while considering cost and latency trade-offs.
Domain 4: Guidelines for Responsible AI
Explores responsible AI development, emphasizing bias detection, fairness, and transparency. It includes using tools like SageMaker Clarify and Bedrock Guardrails to mitigate legal risks and hallucinations, while considering the environmental sustainability of model selection.
Question notes
Multiple choice, multiple response, ordering, and matching questions. No penalty for guessing.
Preparation tips
Familiarize yourself with AWS responsible AI features and practice using SageMaker Model Cards and human-in-the-loop services like Amazon A2I for bias monitoring.
Domain 5: Security, Compliance, and Governance for AI Solutions
Focuses on securing AI workloads through IAM, encryption, and threat detection. It covers governance frameworks, data lineage, and compliance services like AWS Config and Audit Manager to ensure transparency, integrity, and regulatory adherence for AI/ML solutions.
Question notes
Multiple choice, multiple response, ordering, and matching questions. No penalty for guessing.
Preparation tips
Review the AWS shared responsibility model as it applies to AI and practice using security services like Amazon Macie and KMS to protect sensitive training data.