AWS AIF-C01 Practice Exam

AWS Certified AI Practitioner AIF-C01

The AWS Certified AI Practitioner certification validates a foundational understanding of artificial intelligence, machine learning, and generative AI concepts on AWS. The exam covers five domains: Fundamentals of AI and ML, Fundamentals of Generative AI, Applications of Foundation Models, Guidelines for Responsible AI, and Security, Compliance, and Governance for AI Solutions, including services like Amazon Bedrock, SageMaker, Amazon Q, Comprehend, Rekognition, and Textract. Candidates are also tested on prompt engineering, model evaluation, responsible-AI practices, and data governance. This certification suits anyone who uses AI solutions on AWS without building or training models.

839 questions in the bank

Question bank updated : questions, answers, explanations and chapters.

Certification Overview

  • Exam name: AIF-C01
  • Vendor: AWS
  • Exam code: AIF-C01
  • Duration: 90 minutes
  • Total questions: about 65
  • Official passing score: 700 on a scale of 100-1,000 (source)
  • Our practice exam pass mark: 70%

Who Should Take This Exam?

Business Analyst, IT Support Specialist, Marketing Professional, Product Manager, Project Manager, Line-of-Business Manager, IT Manager, Sales Professional, Solutions Architect

Prerequisites

No prior certification is required. AWS recommends up to six months of exposure to AI/ML technologies on AWS.

Topics Covered

  • Fundamentals of AI and ML
  • Fundamentals of GenAI
  • Applications of Foundation Models
  • Guidelines for Responsible AI
  • Security, Compliance, and Governance for AI Solutions

Question Types

  • Multiple Choice (Single Answer)
  • Multiple Choice (Multiple Answers)
  • Drag and Drop

AIF-C01 Practice Questions

Our question bank contains 839+ practice questions for this certification. Sample questions from each exam chapter. Expand a question to see the answer choices. With full access, you get unlimited practice exams with randomized questions from our full question bank.

Fundamentals of AI and ML

What does the term bias mean in the context of a machine learning model?
  • A. Systematic errors that favor certain groups over others Correct answer
  • B. The number of parameters in a neural network
  • C. The cost of running inference at scale
  • D. The time a model takes to generate a response

Option A is correct because bias in a machine learning model refers to systematic errors that cause outputs to favor or disadvantage certain groups over others. Option B describes model size, Option C describes inference cost, and Option D describes latency, none of which relate to fairness or systematic skew.

What is the term for data that has been labeled with the correct output for use in training a model?
  • A. Structured data
  • B. Labeled data Correct answer
  • C. Synthetic data
  • D. Feature data

Option B is correct because labeled data is data paired with the correct output (the ground-truth answer) used to train supervised models. Structured data (A) just refers to tabular/organized data regardless of labels, synthetic data (C) is artificially generated data, and feature data (D) is not a standard term for this concept.

Which type of machine learning uses labeled data to train a model to predict outcomes?
  • A. Unsupervised learning
  • B. Supervised learning Correct answer
  • C. Reinforcement learning
  • D. Semi-supervised learning

Option B is correct because supervised learning trains a model on labeled examples (input paired with known output) to predict outcomes on new data. Unsupervised learning (A) uses unlabeled data to find patterns, reinforcement learning (C) learns from rewards/penalties through interaction, and semi-supervised learning (D) uses a mix but is not the primary answer here.

Which TWO of the following are examples of supervised learning tasks? (Choose two.)
  • A. Association rule mining
  • B. Dimensionality reduction
  • C. Clustering
  • D. Classification Correct answer
  • E. Regression Correct answer

Options D and E are correct because classification and regression both use labeled data to predict a category or numeric value, the definition of supervised learning. Association rule mining (A), dimensionality reduction (B), and clustering (C) are unsupervised techniques that do not require labeled outputs.

Fundamentals of GenAI

What is a foundation model in generative AI?
  • A. A rule-based system with fixed logic
  • B. A dataset labeled for one narrow task
  • C. A large pretrained model adaptable to many tasks Correct answer
  • D. A cluster of GPUs reserved for training

Option C is correct because a foundation model is a large model pretrained on broad data that can be adapted to many downstream tasks through prompting or fine-tuning. Option A describes a rule-based system, which has no learned parameters. Option B describes a narrow, task-specific labeled dataset, not a pretrained model, and option D describes hardware, not a model.

Which AWS service lets a company build a generative AI assistant grounded in its own enterprise documents without managing infrastructure?
  • A. Amazon Kinesis
  • B. AWS Glue
  • C. Amazon Redshift
  • D. Amazon Quick Correct answer

Option D is correct because Amazon Quick is a fully managed generative AI assistant that can be connected to and grounded in a company's own enterprise documents and data sources without the company managing infrastructure. Kinesis, Glue, and Redshift are streaming, ETL, and data warehouse services respectively, not generative AI assistants.

Which AWS service provides access to foundation models from multiple providers through a single API?
  • A. Amazon EMR
  • B. Amazon Bedrock Correct answer
  • C. AWS Lambda
  • D. AWS Batch

Option B is correct because Amazon Bedrock is the fully managed service that exposes foundation models from multiple providers (such as Anthropic, Meta, and Amazon) through a single unified API. EMR, Lambda, and Batch are compute and analytics services with no foundation model access layer.

Which TWO of the following describe a hallucination produced by a generative AI model?
  • A. A confident statement that is factually wrong Correct answer
  • B. An out-of-memory error during training
  • C. A failed API authentication attempt
  • D. A slow response due to network latency
  • E. A fabricated citation that does not exist Correct answer

Options A and E are correct because a hallucination is a confidently stated but factually wrong output, and a fabricated citation that does not exist is a common example of this. Options B, C, and D describe memory errors, authentication failures, and network latency, which are technical or operational issues unrelated to hallucination.

Applications of Foundation Models

Which AWS service is used to customize a foundation model with a company's own labeled examples through a managed fine-tuning job?
  • A. Fine-tune the foundation model in Amazon Bedrock Correct answer
  • B. Convert the examples into an ONNX model
  • C. Package the model as a container image
  • D. Retrain the model from scratch using custom training code

Option A is correct because Amazon Bedrock provides managed fine-tuning jobs that customize a foundation model using a company's own labeled examples without managing training infrastructure. Converting to ONNX, packaging as a container, and retraining from scratch with custom code all involve infrastructure and effort well beyond a managed fine-tuning job.

What does the term prompt mean in generative AI?
  • A. The output text returned by a foundation model
  • B. The file that defines a model's architecture
  • C. The input text or instruction given to a foundation model Correct answer
  • D. The billing metric used to charge for inference

Option C is correct because a prompt is the input text or instruction a user provides to a foundation model to elicit a response. The output text, the model architecture file, and the billing metric are all distinct concepts unrelated to the input given to the model.

Which AWS capability provides vector search for storing embeddings used in retrieval-augmented generation?
  • A. AWS Step Functions
  • B. Amazon Kinesis Data Streams
  • C. Amazon OpenSearch Serverless vector engine Correct answer
  • D. AWS Glue

Option C is correct because the Amazon OpenSearch Serverless vector engine stores embeddings and performs vector similarity search, which is exactly what retrieval-augmented generation needs to retrieve relevant context. Step Functions, Kinesis Data Streams, and Glue are orchestration, streaming, and ETL services with no vector search capability.

A company wants its generative AI application to answer questions using only its own private documents. Which TWO approaches address this requirement? (Choose two.)
  • A. Retrieval Augmented Generation with a knowledge base Correct answer
  • B. Enabling multi-factor authentication on the model endpoint
  • C. Increasing the model's temperature setting
  • D. Fine-tuning the model on the company's documents Correct answer
  • E. Reducing the model's maximum token limit

Options A and D are correct because both RAG with a knowledge base and fine-tuning on company documents let a model draw on private data. Option B is a security control unrelated to data sourcing, Option C affects randomness, and Option E limits response length, neither of which restricts answers to private documents.

Guidelines for Responsible AI

Which AWS service helps detect bias in datasets and trained models?
  • A. Amazon CloudWatch
  • B. AWS Trusted Advisor
  • C. Amazon Kinesis
  • D. Amazon SageMaker Clarify Correct answer

Option D is correct because Amazon SageMaker Clarify is purpose-built to detect statistical bias in datasets and trained models and to help explain model predictions. CloudWatch monitors metrics and logs, Trusted Advisor checks account best practices, and Kinesis streams data, none of which analyze data or models for bias.

A company wants to give customers a plain-language summary of what an AWS AI service does and its intended use cases. Which resource should they consult?
  • A. A CloudTrail log
  • B. An AWS Config rule
  • C. An AWS AI Service Card Correct answer
  • D. A Cost and Usage Report

Option C is correct because an AWS AI Service Card gives customers a plain-language summary of an AWS AI service's intended uses, limitations, and responsible AI considerations. CloudTrail logs API activity, Config rules check resource compliance, and Cost and Usage Reports track spending, none of which describe a service's purpose.

What does explainability refer to in responsible AI?
  • A. The number of AWS Regions a model is deployed in
  • B. The ability to describe how a model reaches its outputs Correct answer
  • C. The speed at which a model can be retrained
  • D. The compression ratio of a stored model

Option B is correct because explainability is the ability to describe or interpret how a model arrives at its outputs. Options A, C, and D describe deployment footprint, retraining speed, and storage efficiency, which have nothing to do with interpreting model behavior.

Which TWO of the following are core dimensions of responsible AI as described by AWS?
  • A. Explainability Correct answer
  • B. Read replica count
  • C. Fairness Correct answer
  • D. Cache eviction policy
  • E. Instance family

Options A and C are correct because explainability and fairness are both core dimensions of AWS's responsible AI approach. Options B, D, and E describe infrastructure settings like replica count, cache eviction, and instance family, which are not responsible AI dimensions.

Security, Compliance, and Governance for AI Solutions

Which AWS service centrally manages encryption keys used to protect data for services such as Amazon SageMaker and Amazon Bedrock?
  • A. AWS Config
  • B. AWS CloudTrail
  • C. AWS KMS Correct answer
  • D. Amazon EventBridge

Option C is correct because AWS KMS is the service that creates, manages, and controls cryptographic keys used to encrypt data for services like SageMaker and Bedrock. AWS Config tracks resource configuration compliance, CloudTrail logs API activity, and EventBridge routes events, none of which manage encryption keys.

What is the purpose of AWS CloudTrail when used with AI services such as Amazon Bedrock?
  • A. Encrypts model artifacts at rest
  • B. Balances inference traffic
  • C. Logs API calls made to the service Correct answer
  • D. Trains foundation models

Option C is correct because AWS CloudTrail records API calls made to AWS services, including who invoked a Bedrock API and when, supporting auditing and governance. It does not encrypt data, balance traffic, or train models, which are unrelated functions handled by other services.

Under the AWS shared responsibility model, which task is AWS responsible for when a company uses a fully managed AI service?
  • A. Writing IAM policies for end users
  • B. Classifying the customer's uploaded data
  • C. Securing the underlying infrastructure Correct answer
  • D. Choosing which prompts to send

Option C is correct because under the shared responsibility model AWS secures the underlying infrastructure, including hardware, host operating systems, and the managed service itself. Writing IAM policies, classifying data, and choosing prompts are customer responsibilities since they involve the customer's own content and access configuration.

Which TWO items would typically appear in a model card?
  • A. Known limitations Correct answer
  • B. Intended use cases Correct answer
  • C. The AWS account billing ID
  • D. The VPC subnet CIDR range
  • E. The IAM role trust policy

Options A and B are correct because known limitations and intended use cases are standard sections of a model card. Options C, D, and E describe billing IDs, subnet ranges, and IAM trust policies, which are account and infrastructure configuration details, not model documentation content.

Free Study Guide

Read the free AWS AIF-C01 study guide: every exam topic explained in plain English, with diagrams, exam tips and the mistakes to avoid.

More AWS Practice Exams

All AWS practice exams →

Practice with realistic mock exams to prepare for your AWS certification.