AWS AI Practitioner Exam Topics, Practice Exams & Test Objectives

AWS Certified AI Practitioner Exam Topics

Format includes 50 scored questions and 15 unscored questions used for calibration
Question types include multiple choice, multiple response, ordering, matching, and case study
Scaled scoring ranges from 100 to 1000 with a passing score of 700
The scoring model is compensatory and you pass based on your overall score
The target candidate has up to six months of exposure to AI and ML on AWS
There is no penalty for guessing so you should answer every question

Explain key terms such as AI, ML, deep learning, neural networks, NLP, training, inference, bias, and LLM
Differentiate supervised, unsupervised, and reinforcement learning and describe batch and real-time inference
Identify data types including labeled, unlabeled, tabular, time series, image, text, structured, and unstructured
Map use cases to techniques such as regression, classification, and clustering
Describe the ML lifecycle including data collection, EDA, preprocessing, feature engineering, training, tuning, evaluation, deployment, and monitoring
Recognize AWS services for each stage such as SageMaker, Data Wrangler, Feature Store, and Model Monitor

Understand tokens, chunking, embeddings, vectors, prompt engineering, transformers, multimodal and diffusion models
Identify use cases including generation, summarization, chat, translation, code, agents, search, and recommendations
Describe the foundation model lifecycle from data selection and pretraining to deployment and feedback
Assess advantages and limitations including adaptability, hallucinations, interpretability, accuracy, and nondeterminism
Evaluate model selection factors including capability, performance, constraints, and compliance and define value metrics such as conversion rate and CLV
Recognize AWS options such as Amazon Bedrock, PartyRock, Amazon Q, and SageMaker JumpStart

Select pre-trained models by cost, modality, latency, multilingual support, size, customization, and context length
Tune inference parameters such as temperature and input or output length to control responses
Define RAG and apply Bedrock knowledge bases with suitable vector stores
Choose embedding storage with OpenSearch Service, Aurora, Neptune, DocumentDB, or RDS for PostgreSQL
Compare customization approaches including pretraining, fine-tuning, in-context learning, and RAG based augmentation
Use agents to orchestrate multi-step tasks with Agents for Amazon Bedrock

Identify pillars of responsible AI such as bias, fairness, inclusivity, robustness, safety, and veracity
Use guardrails and tooling including Guardrails for Amazon Bedrock, SageMaker Clarify, Model Monitor, and Amazon A2I
Consider sustainability and environmental factors when selecting models
Recognize legal risks including IP issues, biased outputs, loss of trust, and hallucinations
Assess dataset characteristics for inclusivity, diversity, balance, and curation quality
Apply human-centered design principles for explainable and transparent systems

Secure AI systems with IAM roles and policies, encryption, Macie, PrivateLink, and shared responsibility awareness
Track data lineage and origins using catalogs and SageMaker Model Cards
Apply secure data engineering practices including privacy enhancing techniques, access control, and integrity checks
Account for threats such as prompt injection and ensure encryption at rest and in transit
Align to regulations and frameworks using AWS Config, Inspector, Audit Manager, Artifact, CloudTrail, and Trusted Advisor
Plan governance with policies, reviews, transparency standards, training, and the Generative AI Security Scoping Matrix

Developing or coding models or algorithms
Implementing data engineering or feature engineering techniques
Hyperparameter tuning or detailed model optimization
Building and deploying AI or ML pipelines and infrastructure
Mathematical or statistical analysis of models or implementing AI security protocols
Creating governance frameworks and policies for AI solutions

Study the official exam guide and the task statements for all five domains
Practice recognizing use cases and matching them to appropriate AWS services
Experiment with Amazon Bedrock, SageMaker JumpStart, and prompt engineering in sandboxes
Review responsible AI topics including guardrails, transparency, and evaluation methods
Do mixed-format practice with multiple choice, multiple response, ordering, matching, and case study items

Get certified with the help of Cameron McKenzie (@cameronmcnz), the industry's foremost expert AI, ML, DevOps and Cloud Computing.