AWSAWS Certified AI Practitioner
The AWS Certified AI Practitioner certification validates your understanding of AI, machine learning, and generative AI concepts and use cases on AWS. It is designed for individuals in business, sales, marketing, and IT roles who work with AI/ML technologies but do not necessarily build them. Earning this credential demonstrates that you can identify the right AI solutions for business problems and apply them responsibly, positioning you for career growth in an AI-driven market.
398 practice questions · Updated 2026-07-30
5Domains
14Objectives
98Concepts
398Questions
AIF-C01 Curriculum
Every domain, objective, and concept the AIF-C01 exam measures.
- Basic AI terminologyDefine key AI terms such as AI, ML, deep learning, neural networks, computer vision, NLP, model, algorithm, training, inferencing, bias, fairness, fit, LLM, GenAI, and agentic AI.
- Relationships among AI, ML, GenAI, deep learning, and agentic AIDescribe the similarities and differences between AI, ML, GenAI, deep learning, and agentic AI, including their scope and relationships.
- Types of inferencingDescribe various types of inferencing, including batch, real-time, asynchronous, and serverless, and their use cases.
- Types of data in AI modelsDescribe different types of data used in AI models, such as labeled, unlabeled, tabular, time-series, image, text, structured, and unstructured.
- Types of AI/ML learningDescribe different types of AI/ML learning methods, including supervised learning, unsupervised learning, and reinforcement learning.
- Value-driven AI/ML use casesIdentify scenarios where AI/ML adds value, such as assisting human decision-making, enabling scalability, and automating tasks.
- Inappropriate AI/ML applicationsDetermine when AI/ML is not suitable, considering cost-benefit trade-offs and the need for deterministic outcomes rather than predictions.
- Matching techniques to use casesSelect appropriate AI/ML techniques (e.g., regression, classification, clustering) based on the specific problem type and data characteristics.
- Real-world AI application examplesRecognize examples of real-world AI applications, including computer vision, NLP, speech recognition, recommendation systems, fraud detection, forecasting, knowledge bases, and agentic AI.
- AWS managed AI/ML services capabilitiesExplain the capabilities of AWS managed AI/ML services such as Amazon SageMaker AI, Transcribe, Translate, Comprehend, Lex, and Polly.
- Traditional ML vs. foundation models selectionIdentify when to use traditional ML models versus foundation models, considering regulatory, explainability, and operational constraints.
- AI/ML pipeline componentsDescribe the stages of an AI/ML pipeline (e.g., data collection, preprocessing, training, evaluation, deployment) and differentiate them.
- Sources of foundation modelsDifferentiate between using open-source pre-trained models, custom-trained models, and other sources of FM models.
- Model deployment methodsDescribe methods to use a model in production, such as managed API services and self-hosted APIs.
- AWS services for AI/ML pipeline stagesIdentify relevant AWS services (e.g., Amazon Bedrock, Amazon Q, Amazon QuickSight, Kiro, SageMaker AI) for each stage of an AI/ML pipeline.
- MLOps fundamental conceptsDescribe fundamental MLOps concepts including experimentation, repeatable processes, scalable systems, managing technical debt, production readiness, model monitoring, and model re-training.
- Model performance metricsDescribe model performance metrics such as accuracy, precision, recall, and F1 score.
- Business metrics for ML modelsDescribe business metrics like cost per user, development costs, customer feedback, and ROI to evaluate ML models.
- Foundational GenAI conceptsDefine tokens, chunking, embeddings, vectors, prompt engineering, transformer-based LLMs, foundation models, multi-modal models, and diffusion models.
- GenAI use casesIdentify potential use cases for GenAI models across image, video, audio, text, code, and conversational applications.
- Foundation model lifecycleDescribe the stages of the FM lifecycle including data selection, model selection, pre-training, fine-tuning, evaluation, deployment, and feedback.
- Token-based pricing and inferenceExplain how token-based pricing works and its impact on cost and performance during inference.
- Context engineeringDescribe the role of context engineering in optimizing FM applications by providing relevant context to improve outputs.
- Agentic AI conceptsDefine foundational agentic AI concepts including multi-agent system patterns, MCP, communication patterns, memory management, tool usage, and workflow orchestration.
- Advantages of GenAIDescribe the advantages of GenAI, including adaptability, responsiveness, conversational capabilities, and content generation.
- Disadvantages of GenAIIdentify disadvantages of GenAI solutions, such as hallucinations, interpretability issues, inaccuracy, and nondeterminism.
- Model selection factorsIdentify factors to consider when selecting GenAI models, including model types, performance, capabilities, constraints, compliance, cost, latency, and complexity.
- Business value and metricsDetermine business value and metrics for GenAI applications, such as ROI, efficiency, conversion rate, average revenue per user, accuracy, and customer lifetime value.
- AWS services for GenAI developmentIdentify AWS services and features such as Amazon Bedrock, Amazon SageMaker AI, SageMaker JumpStart, Amazon Quick, Kiro, Strands Agents, and Amazon Bedrock AgentCore used for building GenAI applications.
- Advantages of AWS GenAI servicesDescribe the advantages of using AWS GenAI services, including accessibility, lower barrier to entry, efficiency, cost-effectiveness, speed to market, and ability to meet business objectives.
- Benefits of AWS infrastructure for GenAIDescribe the benefits of AWS infrastructure for GenAI applications, such as security, compliance, responsibility, and safety.
- Cost tradeoffs in AWS GenAI servicesDescribe cost tradeoffs of AWS GenAI services, including responsiveness, availability, redundancy, performance, regional coverage, token-based pricing, provisioned throughput, and custom models.
- FM selection criteriaIdentify and apply criteria such as cost, modality, latency, multilingual support, model size, complexity, customization, input/output length, and prompt caching when choosing a foundation model.
- Inference parametersExplain how inference parameters like temperature and input/output length affect model responses and adjust them for desired outcomes.
- Retrieval Augmented Generation (RAG)Define RAG and describe its business applications, including how Amazon Bedrock Knowledge Bases implement it.
- Vector databases in AWSIdentify AWS services that store embeddings, such as Amazon OpenSearch Service, Amazon Aurora, Amazon Neptune, and Amazon RDS for PostgreSQL.
- FM customization cost tradeoffsCompare the cost tradeoffs of pre-training, fine-tuning, in-context learning, RAG, and model distillation for customizing foundation models.
- AI agentsDefine the role of AI agents and describe their business applications.
- Core prompt engineering constructsDefine and differentiate context, instruction, and negative prompts, and explain how each shapes model output.
- Prompt engineering techniquesDescribe chain-of-thought, zero-shot, single-shot, few-shot, and prompt templates, and identify when to use each.
- Benefits and best practicesIdentify benefits like improved response quality and describe best practices including experimentation, guardrails, discovery, specificity, concision, and using multiple comments.
- Risks and limitationsDefine exposure, poisoning, hijacking, and jailbreaking, and explain how they threaten prompt-based systems.
- Prompt versioning and managementDescribe strategies for versioning and managing prompts using Amazon Bedrock Prompt Management.
- Pre-trainingDescribe the purpose and process of pre-training a foundation model on large unlabeled datasets to learn general language patterns.
- Fine-tuningExplain how fine-tuning adapts a pre-trained model to a specific task or domain by training on a smaller labeled dataset.
- Continuous pre-trainingDescribe continuous pre-training as a method to further train a model on domain-specific data to improve its performance in that domain.
- DistillationDefine knowledge distillation as a technique to transfer knowledge from a large model to a smaller one, preserving performance while reducing size.
- Instruction tuningExplain instruction tuning, which fine-tunes a model on examples of instructions and desired responses to improve its ability to follow user prompts.
- Transfer learningDescribe transfer learning as leveraging knowledge from a pre-trained model to a new task, reducing the need for extensive training data.
- Data curationExplain the importance of curating high-quality, relevant data for fine-tuning, including cleaning, filtering, and organizing datasets.
- Data governanceDescribe data governance considerations in fine-tuning, such as ensuring compliance with regulations, privacy, and ethical use of data.
- Data size and labelingExplain how the size and labeling of the fine-tuning dataset affect model performance, including the need for sufficient and accurately labeled examples.
- Data representativenessDescribe the importance of ensuring the fine-tuning data is representative of the target domain and user population to avoid bias and improve generalization.
- Reinforcement learning from human feedback (RLHF)Define RLHF as a fine-tuning method that uses human feedback to optimize model behavior, aligning outputs with human preferences.
- Human-in-the-loop evaluationDescribe how human evaluators assess FM outputs for quality, safety, and relevance.
- Benchmark datasetsIdentify standard datasets used to evaluate FM performance across tasks like language understanding and generation.
- Amazon Bedrock Model EvaluationExplain how Amazon Bedrock's Model Evaluation feature automates evaluation using built-in or custom metrics.
- ROUGE metricDefine ROUGE and its use in evaluating text summarization by comparing n-gram overlap.
- BLEU metricDefine BLEU and its use in evaluating machine translation by measuring precision of n-gram matches.
- BERTScore metricExplain BERTScore as a semantic similarity metric that uses contextual embeddings to evaluate text generation.
- LLM-as-a-judgeDescribe using a large language model to score or compare outputs of other models.
- Business objective alignmentDetermine whether an FM meets business goals such as productivity, user engagement, and task efficiency.
- RAG evaluationIdentify methods to evaluate retrieval-augmented generation systems, including retrieval quality and answer faithfulness.
- Agent evaluationDescribe approaches to evaluate agent performance, such as task success rate and tool use correctness.
- Workflow evaluationExplain how to evaluate multi-step workflows built with FMs, focusing on end-to-end outcomes and error rates.
- Task completion rateDefine task completion rate as a metric measuring the percentage of user tasks successfully completed.
- User satisfactionExplain user satisfaction metrics like surveys or ratings to assess AI application experience.
- Cost per interactionDescribe cost per interaction as a business metric evaluating the financial efficiency of AI applications.
- Features of Responsible AIIdentify and describe the key features of responsible AI, including bias, fairness, inclusivity, robustness, safety, and veracity.
- Tools for Responsible AI FeaturesExplain how to use tools such as Amazon Bedrock Guardrails to identify and enforce features of responsible AI in AI systems.
- Responsible Model Selection PracticesDefine responsible practices for selecting AI models, considering environmental impact and sustainability.
- Legal Risks in Generative AIIdentify legal risks associated with generative AI, including intellectual property infringement, biased outputs, loss of customer trust, end-user risk, and hallucinations.
- Characteristics of Responsible DatasetsIdentify characteristics of datasets that support responsible AI, such as inclusivity, diversity, curated sources, and balance.
- Effects of Bias and VarianceDescribe the effects of bias and variance in AI models, including impacts on demographic groups, inaccuracy, overfitting, and underfitting.
- Tools for Detecting and Monitoring Bias and TrustworthinessDescribe tools and methods to detect and monitor bias, trustworthiness, and truthfulness, including label quality analysis, human audits, subgroup analysis, Amazon SageMaker Clarify, SageMaker Model Monitor, and Amazon A2I.
- Transparent vs. explainable modelsDescribe the differences between models that are transparent and explainable and those that are not, including characteristics like interpretability and opacity.
- Tools for transparency and explainabilityIdentify tools such as Amazon SageMaker Model Cards, SageMaker Clarify, Amazon Bedrock Model Evaluations, and open source models, data, and licensing considerations for assessing model transparency.
- Tradeoffs between safety and transparencyIdentify tradeoffs between model safety and transparency, including how interpretability measures may affect performance and safety.
- Human-centered design for explainable AIDescribe principles of human-centered design for explainable AI, including user-feedback mechanisms and AI decision transparency.
- AWS shared responsibility modelExplain how security responsibilities are divided between AWS and the customer for AI services.
- IAM roles, policies, and permissions for AI servicesDescribe how to use IAM to control access to AI resources and actions.
- Encryption for AI dataIdentify encryption options (at rest and in transit) for AI workloads and data.
- Amazon Macie for sensitive data discoveryDescribe how Amazon Macie helps discover and protect sensitive data in AI pipelines.
- AWS PrivateLink for private connectivityExplain how AWS PrivateLink provides private, secure access to AI services.
- Amazon Bedrock AgentCore Identity and PolicyDescribe how AgentCore identity and policies manage permissions for Bedrock agents.
- Amazon Bedrock GuardrailsExplain how Bedrock Guardrails enforce safety and security policies for generative AI applications.
- Data lineageDefine data lineage and its role in tracking data origins and transformations.
- Data catalogingDescribe how data cataloging helps document and manage data assets for AI.
- Amazon SageMaker Model CardsExplain how SageMaker Model Cards document model provenance, intended use, and evaluation details.
- Secure data engineering best practicesList best practices for securely ingesting, storing, and processing data for AI systems.
- AI Governance FrameworksIdentify common governance frameworks and principles for responsible AI, such as those from NIST, OECD, or ISO, and their key components.
- Regulatory Landscape for AIRecognize major AI-related regulations and standards (e.g., EU AI Act, GDPR, sector-specific rules) and their implications for AI systems.
- Compliance Requirements for AIDetermine compliance obligations for AI systems, including data protection, privacy, and industry-specific mandates.
- AI Risk ManagementApply risk management concepts to AI systems, including risk assessment, mitigation strategies, and continuous monitoring.
- Data Governance in AIUnderstand data governance practices relevant to AI, such as data quality, lineage, and lifecycle management, to ensure compliance.
- Model Governance and DocumentationRecognize the importance of model governance, including documentation, versioning, and audit trails for AI models.
- Ethical Considerations in AIIdentify ethical principles (fairness, transparency, accountability) and how they align with governance and compliance.
- Auditing and Reporting for AIDescribe processes for auditing AI systems and reporting compliance status to stakeholders and regulators.
Ready to practice?Test your knowledge with exam-style questions or take an intelligent quiz tailored to your level.
Percentages reflect share of the current practice bank, not official exam weightings — no structured per-skill weight is published for AIF-C01, so none is invented.