A W S   C E R T I F I C A T I O N

AIF C01 Certified AI Practitioner Practice Exam

Exam Number: 1202 | Last updated April 24, 2026 | 700+ questions across 4 vendor-aligned objectives

The AWS Certified AIF C01 AI Practitioner targets professionals who use, evaluate, or sell AI and generative AI solutions on AWS without necessarily building the underlying models. Typical candidates include business analysts, product managers, sales engineers, IT support engineers, and technologists pivoting toward AI-adjacent roles. The exam tests the vocabulary, service catalog, and ethical guardrails needed to participate credibly in modern AI conversations.

Fundamentals of Generative AI and Applications of Foundation Models together account for the largest share of the exam. Fundamentals of Generative AI (24%) covers tokens, embeddings, prompt engineering basics, and the generative AI lifecycle. Applications of Foundation Models (28%) is the heaviest weighted domain and addresses Amazon Bedrock, Retrieval Augmented Generation (RAG) with Amazon Kendra and OpenSearch, model selection trade-offs, and Amazon Q for business and developer use cases.

The remaining domains round out the breadth of an AI-aware practitioner. Fundamentals of AI and ML (20%) covers supervised, unsupervised, and reinforcement learning, plus core AWS services such as Amazon SageMaker, Amazon Rekognition, Amazon Comprehend, and Amazon Transcribe. Guidelines for Responsible AI (14%) covers bias, fairness, explainability, and Amazon SageMaker Clarify. Security, Compliance, and Governance for AI Solutions (14%) covers AWS Identity and Access Management for AI workloads, model evaluation, and AWS compliance frameworks.

♥  Get comfortable with the differences between Amazon Bedrock, Amazon SageMaker, and Amazon Q before exam day — many distractor answers swap one service for another. Practice articulating when RAG is appropriate versus fine-tuning, since this trade-off appears in multiple scenarios. Memorize which Amazon AI service does which modality (Rekognition for vision, Transcribe for speech-to-text, Polly for text-to-speech, Comprehend for NLP).

Every answer links to the source. Each explanation below includes a hyperlink to the exact AWS documentation page the question was derived from. PowerKram is the only practice platform with source-verified explanations. Learn about our methodology →

489

practice exam users

95.5%

satisfied users

455%

passed the exam

4.4/5

quality rating

Test your AIF C01 Certified AI Practitioner knowledge

10 of 700+ questions

Question #1 - Fundamentals of AI and ML

A retail company wants a model that can identify product categories from photos uploaded by users.

What type of machine learning task is this?

A) Regression
B) Reinforcement learning
C) Supervised classification
D) Unsupervised clustering

 

Correct answers: C – Explanation:
Mapping images to a known set of categories using labeled training data is supervised classification. Regression predicts continuous values; reinforcement learning learns from reward signals; unsupervised clustering groups data without labels. Source: [Types of ML Problems](https://docs.aws.amazon.com/machine-learning/latest/dg/types-of-ml-models.html)

A marketing team wants to use a foundation model to generate product descriptions but does not want to train a model from scratch.

Which AWS service lets them access multiple foundation models through a single API?

A) Amazon SageMaker Training Jobs
B) AWS Glue
C) Amazon Comprehend
D) Amazon Bedrock

 

Correct answers: D – Explanation:
Amazon Bedrock provides a unified API to access foundation models from Anthropic, Amazon, Meta, AI21, Cohere, Stability AI, and Mistral without managing infrastructure. SageMaker Training is for training custom models; Comprehend is a pre-built NLP service; Glue is ETL. Source: [What is Amazon Bedrock](https://docs.aws.amazon.com/bedrock/latest/userguide/what-is-bedrock.html)

A team wants a chatbot to answer questions using their internal product documentation, but they do not want to fine-tune a model.

Which technique best fits this requirement?

A) Pre-training a new foundation model
B) Retrieval-Augmented Generation (RAG)
C) Reinforcement learning from human feedback
D) Continued pretraining on the entire internet

 

Correct answers: B – Explanation:
RAG retrieves relevant documents at inference time and supplies them to the model as context, letting the model answer from your data without retraining or fine-tuning. Pre-training and continued pretraining require massive compute; RLHF is for aligning model behavior, not adding knowledge. Source: [What is RAG](https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base.html)

An ML engineer notices that a loan-approval model approves applicants from one zip code at a much higher rate than from a demographically similar adjacent zip code.

Which responsible-AI concern is most directly raised, and which AWS tool helps detect it?

A) Hallucination — Amazon Bedrock Guardrails
B) Cost — AWS Cost Explorer
C) Latency — Amazon CloudWatch
D) Bias — Amazon SageMaker Clarify

 

Correct answers: D – Explanation:
Disparate approval rates across similar groups indicate bias. SageMaker Clarify detects bias before, during, and after training and provides explainability. Hallucination is about fabricated outputs; latency and cost are not fairness concerns. Source: [Amazon SageMaker Clarify](https://docs.aws.amazon.com/sagemaker/latest/dg/clarify-fairness-and-explainability.html)

A developer wants to constrain a Bedrock-hosted model so it refuses to discuss competitor products and never returns profanity.

Which Bedrock feature implements this directly?

A) Bedrock Guardrails
B) Provisioned Throughput
C) Bedrock Agents
D) Bedrock Knowledge Bases

 

Correct answers: A – Explanation:
Guardrails apply configurable content filters, denied topics, and word/phrase blocks to model inputs and outputs. Provisioned Throughput is a capacity model; Agents orchestrate multi-step tasks; Knowledge Bases ground answers in private data. Source: [Amazon Bedrock Guardrails](https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails.html)

A regulated financial firm wants foundation-model API traffic to never traverse the public internet.

Which network feature should they use?

A) Public endpoint with TLS
B) Direct Connect to amazonaws.com
C) NAT Gateway in a private subnet
D) AWS PrivateLink (VPC endpoint) for Amazon Bedrock

 

Correct answers: D – Explanation:
PrivateLink/VPC endpoints let your VPC reach Bedrock without going over the public internet, satisfying the requirement. TLS over the public endpoint still uses the internet; NAT Gateway routes outbound to the internet; Direct Connect is for hybrid links, not service privacy. Source: [Bedrock VPC endpoints](https://docs.aws.amazon.com/bedrock/latest/userguide/vpc-interface-endpoints.html)

An engineer wants to provide a few example input/output pairs in the prompt so the model learns the desired format without being retrained.

What is this technique called?

A) Fine-tuning
B) Continued pretraining
C) Few-shot prompting
D) Distillation

 

Correct answers: C – Explanation:
Few-shot prompting embeds a small number of examples directly in the prompt to steer the model. Fine-tuning, continued pretraining, and distillation all change model weights; few-shot does not. Source: [Prompt engineering guidelines](https://docs.aws.amazon.com/bedrock/latest/userguide/prompt-engineering-guidelines.html)

A company has 5,000 internal policy documents and wants users to query them in natural language with citations to the source document.

Which managed feature provides this end-to-end?

A) Amazon Bedrock Knowledge Bases
B) Amazon Polly
C) Amazon Translate
D) Amazon Rekognition

 

Correct answers: A – Explanation:
Bedrock Knowledge Bases ingests documents, chunks and embeds them into a vector store, and returns grounded answers with source citations. Polly is text-to-speech, Translate is translation, Rekognition is image/video analysis. Source: [Bedrock Knowledge Bases](https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base.html)

A model trained on historical hiring data starts to recommend candidates that mirror past biased hiring patterns.

What is the underlying issue called?

A) Overfitting on validation data
B) Training data bias
C) Catastrophic forgetting
D) Prompt injection

 

Correct answers: B – Explanation:
When training data reflects historical biases, the model encodes and reproduces them. Overfitting is memorizing training examples; catastrophic forgetting is losing prior knowledge during fine-tuning; prompt injection is an inference-time attack. Source: [Responsible use of ML](https://docs.aws.amazon.com/sagemaker/latest/dg/clarify-fairness-and-explainability.html)

A team wants to choose between a small, fast model and a large, more accurate one for a high-volume customer-support summarization use case.

Which trade-off is most relevant?

A) Accuracy vs. inference cost and latency
B) Storage vs. networking
C) On-premises vs. hybrid
D) Open-source vs. commercial license only

 

Correct answers: A – Explanation:
Larger models are typically more accurate but cost more per token and respond more slowly — a critical trade-off at high volume. The other pairs are not the central concern of model selection. Source: [Choosing a foundation model](https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html)

Get 700+ more questions with source-linked explanations

Every answer traces to the exact AWS documentation page — so you learn from the source, not just memorize answers.

Start free 24-hour trial — no credit card

Exam mode & learn mode · Score by objective · Updated April 24, 2026

Learn more...

What the aws-ai-practitioner exam measures

  • Fundamentals of AI and ML (20%) — Identify ML lifecycle stages, recognize supervised, unsupervised, and reinforcement learning, and select appropriate AWS AI services for common business problems.
  • Fundamentals of Generative AI (24%) — Define foundation models, tokens, embeddings, and prompt engineering principles; explain the generative AI development lifecycle on AWS.
  • Applications of Foundation Models (28%) — Choose appropriate model and inference parameters in Amazon Bedrock, design RAG architectures, and evaluate Amazon Q use cases for business and code generation.
  • Guidelines for Responsible AI (14%) — Apply responsible-AI principles, identify bias and fairness concerns, and use Amazon SageMaker Clarify and Bedrock Guardrails to mitigate risk.
  • Security, Compliance, and Governance for AI Solutions (14%) — Apply AWS Identity and Access Management to AI workloads, govern model deployments, and align AI use with AWS compliance frameworks.

  • Review the official AWS exam guide and confirm the latest domain weights and content scope before scheduling.
  • Complete the matching learning plan on AWS Skill Builder, including the digital courses and exam prep modules.
  • Build hands-on muscle memory in an AWS Free Tier account by deploying the services that appear in the Applications of Foundation Models domain.
  • Apply your skills to a real-world project — workplace assignments, volunteer work, or open-source contributions where AWS services solve a concrete problem.
  • Master one objective at a time, beginning with the highest-weighted domain so the score impact of each study session is maximized.
  • Run PowerKram in Learn mode to read the explanations and follow every sourced documentation link until you can predict the right answer before reading the choices.
  • Switch to PowerKram Exam mode across all objectives once your accuracy in Learn mode passes 85%, simulating the timed exam experience.

Holders of the AI Practitioner credential typically pivot toward AI-adjacent roles or strengthen their position in current ones:

Anchor your study in AWS-authored material, then drill repetition with PowerKram:

Related certifications to explore

MLS-C01 Machine Learning Specialty
MLE-C01 ML Engineer Associate
CLF-C02 Cloud Practitioner
AI-900 Azure AI Fundamentals
Google Generative AI Leader
AI-102 Azure AI Engineer

Related reading from our Learning Hub

AWS AI and Machine Learning
PowerKram vs ExamTopics
Responsible AI and Ethics
PowerKram vs IT-Exams