SaveMyCert
Log in
5 of 5 free questions left today·for 10 a day
AIF-C01 · Domain 2

Fundamentals of GenAI practice questions

Fundamentals of GenAI is worth 24% of the AIF-C01 exam — the 2nd-heaviest of the 5 domains. Generative AI concepts, its capabilities and limitations for business problems, and the AWS infrastructure for building GenAI applications. 6 fully worked examples are further down this page, answers included.

Exam weight
24%
the 2nd-heaviest of the 5 domains
Questions
102
across 3 topics
Free, no account
5/day
a free account allows 10 a day
Explanations
Every option
right and wrong

Build a practice session

5 free questions left today.

Domains

How many?

5 free questions left today.

Mode

Ready when you are

10 fresh questions drawn across 1 of 5 domains, in Learn mode.

Review what you missed

Every question you answer incorrectly, and every question you flag while practising, is saved here automatically. Finish a session and you can come back to re-drill just those.

6 sample Fundamentals of GenAI questions, fully explained

Questions from the AIF-C01 bank mapped to domain 2, with the answer key and the reasoning behind every option. None of them repeat the examples on the main AIF-C01 practice page.

Question 1Fundamentals of GenAI

A company wants to make a 200-page product manual searchable by a generative AI application. It first splits the manual into smaller sections of a few paragraphs each. What is this practice called?

Choose one.

  • a
    Tokenizing

    Tokenizing breaks text into the tiny units a model reads, not a document into paragraph-sized sections.

  • b
    Chunking Correct

    Chunking splits a large document into smaller, manageable pieces for processing or retrieval.

  • c
    Fine-tuning

    Fine-tuning further trains a model on task-specific data; it does not divide a document.

  • d
    Embedding

    Embedding turns content into numbers that capture meaning; it does not split a document into sections.

The concept

Chunking means dividing a large document into smaller pieces so an application can store, search and retrieve the relevant parts.

Why that’s the answer

A 200-page manual does not fit in one context window, so the application splits it into paragraph-sized sections and later pulls only the relevant ones into a prompt. That is chunking. Tokenizing also splits text, but into tiny word-level units that the model reads; fine-tuning further trains a model; and embedding converts each piece into numbers that capture its meaning, which usually happens after chunking.

How to reason it out
  1. Notice what is being split: a whole document into sections of a few paragraphs.
  2. Match document-level splitting for retrieval to chunking.
  3. Separate it from tokenizing (word-level units), embedding (numbers for meaning) and fine-tuning (training).

Exam tip: Chunking splits a large document into sections an application can handle; tokenizing splits text into the units a model reads.

Generative AI Concepts: Tokens, Embeddings, and Foundation Models (AIF-C01) — the lesson that teaches this.

Question 2Fundamentals of GenAI

Which statement best describes an embedding in generative AI?

Choose one.

  • a
    The small units of text that a language model reads and writes one at a time.

    Those are tokens. An embedding is produced from content, it is not the text units themselves.

  • b
    The maximum number of tokens that a model can consider within a single request.

    That is the context window, a size limit, not a representation of meaning.

  • c
    A numerical representation of content in which similar meanings get similar values. Correct

    Correct. An embedding encodes the meaning of a word, sentence or document as numbers, so similar meanings land close together.

  • d
    A ranked list of the documents that a keyword search index returns for a user's query.

    Keyword search results are matched on words, not meaning, and are not an embedding.

The concept

An embedding is a numerical representation of a piece of content, produced by a model so that content with similar meaning gets similar numbers.

Why that’s the answer

Embeddings let applications compare meaning by comparing numbers, which is what powers semantic search and recommendations. Tokens are the text units a model processes, the context window is the token limit of a request, and a ranked keyword-search result list is matched on words rather than meaning.

How to reason it out
  1. Look for the description that turns meaning into numbers.
  2. Recall that similar meanings produce similar embeddings, expressed as vectors.
  3. Eliminate tokens, the context window and keyword search results as different concepts.

Exam tip: An embedding encodes meaning as numbers, so similar meanings map to similar values.

Generative AI Concepts: Tokens, Embeddings, and Foundation Models (AIF-C01) — the lesson that teaches this.

Question 3Fundamentals of GenAI

An HR portal's keyword search finds nothing when employees ask "How do I take time off?" because the relevant policy page is titled "Leave entitlement". The team now embeds every policy page and each incoming question. How does the new search decide which page is relevant?

Choose one.

  • a
    It ranks pages by how many words their stored text has in common with the question.

    Counting shared words is keyword matching, which is exactly what failed here: the question and the page use different words.

  • b
    It ranks pages whose vectors share the most identical numbers with the question's vector.

    Related meanings give nearby vectors, not matching numbers position by position; similarity is measured as distance or angle between whole vectors.

  • c
    It ranks pages whose vectors sit closest to the question's vector in the embedding space. Correct

    Correct. Embeddings place similar meanings near each other, so the nearest vectors are the most semantically similar pages, even with no shared words.

  • d
    It converts the question's vector back into keywords and reruns the keyword search.

    Semantic search compares vectors directly. Turning the question back into keywords would bring back the vocabulary mismatch that caused the failure.

The concept

A vector is the list of numbers an embedding produces, and distance between vectors reflects similarity of meaning.

Why that’s the answer

Because embeddings place similar meanings at nearby points, the page about leave entitlement lands close to a question about taking time off even though they share no keywords, and the system returns the nearest vectors. Counting shared words is the keyword search that already failed, and converting the vector back into keywords repeats that failure. Similar meanings produce vectors that are close overall, not vectors with identical numbers, so counting matching values is not how similarity is measured.

How to reason it out
  1. Notice the failure: the question and the page mean the same thing but use different words.
  2. Recall that embeddings represent meaning as vectors, and similar meanings sit close together.
  3. Conclude that the nearest vectors are returned, which is semantic search.

Exam tip: Semantic search returns the content whose vectors are closest to the query's vector, so it matches meaning rather than keywords.

Generative AI Concepts: Tokens, Embeddings, and Foundation Models (AIF-C01) — the lesson that teaches this.

Question 4Fundamentals of GenAI

A marketing team wants a model that produces original product images from short text descriptions. Which type of generative model is designed mainly for this task?

Choose one.

  • a
    A large language model (LLM)

    An LLM generates text, so it cannot produce the images the team needs.

  • b
    A text embedding model

    An embedding model turns text into vectors for search and comparison; it does not generate images.

  • c
    A speech-to-text model

    A speech-to-text model transcribes audio into text; it does not create pictures.

  • d
    A diffusion model Correct

    Correct. Diffusion models are the model type used mainly for image generation, refining random noise into an image that matches the prompt.

The concept

A diffusion model is a generative model used mainly for image generation from a prompt.

Why that’s the answer

Diffusion models start from random noise and refine it step by step into an image that matches the text description, which is the team's need. An LLM outputs text, an embedding model outputs vectors for comparison, and a speech-to-text model outputs transcripts, so none of them produces images.

How to reason it out
  1. Identify the output: new images created from text.
  2. Match image generation from a prompt to the diffusion model.
  3. Rule out models whose output is text, vectors or transcripts.

Exam tip: Diffusion models generate images by refining noise into a picture that matches the prompt.

Generative AI Concepts: Tokens, Embeddings, and Foundation Models (AIF-C01) — the lesson that teaches this.

Question 5Fundamentals of GenAI

An insurance app lets a customer upload a photo of a damaged car and type a question about it. The app must return a written answer that uses both the photo and the question. What type of model fits this requirement?

Choose one.

  • a
    A large language model trained on text

    A model that works on text can read the question but cannot interpret the photo.

  • b
    A diffusion model

    A diffusion model generates images; it is not designed to read an image and answer a question about it in text.

  • c
    A text embedding model

    An embedding model turns text into vectors for comparison; it neither reads the photo nor writes an answer.

  • d
    A multi-modal model Correct

    Correct. A multi-modal model accepts more than one kind of data, such as an image plus text, and can produce a text answer.

The concept

A multi-modal model handles more than one type of data, such as images and text, in a single model.

Why that’s the answer

The input mixes an image with a text question, so the model must understand both modalities at once. A text-trained LLM cannot see the photo, a diffusion model is built to generate images rather than to answer questions about them, and an embedding model produces vectors, not answers.

How to reason it out
  1. Note that the input combines two data types: an image and text.
  2. Match several data types in one model to a multi-modal model.
  3. Rule out text-trained, image-generation and embedding models.

Exam tip: A multi-modal model works with several kinds of data, such as text and images, at once.

Generative AI Concepts: Tokens, Embeddings, and Foundation Models (AIF-C01) — the lesson that teaches this.

Question 6Fundamentals of GenAI

What is a foundation model (FM)?

Choose one.

  • a
    A compact model trained from scratch to perform one narrow task very well.

    A foundation model is the opposite: large, general and reused across many tasks rather than built for one job.

  • b
    A large model pre-trained on broad data that can be adapted to many tasks. Correct

    Correct. Broad pre-training and adaptability to many downstream tasks are the defining traits of a foundation model.

  • c
    A rule-based system that applies fixed, hand-written logic to each request.

    A foundation model learns patterns from data; it does not follow hand-written rules.

  • d
    A vector database that stores embeddings so that applications can search them.

    That describes a vector store used for semantic search, not a model.

The concept

A foundation model (FM) is a large, general-purpose model pre-trained on broad data and adaptable to many tasks.

Why that’s the answer

The defining traits are breadth of pre-training and reuse across tasks such as answering questions, summarizing and translating; an LLM is a foundation model specialized for text. A narrow single-task model, a hand-written rule system and a vector database each miss that general, pre-trained nature.

How to reason it out
  1. Recall that a foundation model is trained broadly, not for one job.
  2. Look for the description combining broad pre-training with adaptation to many tasks.
  3. Rule out the narrow, rule-based and storage descriptions.

Exam tip: A foundation model is broadly pre-trained and can be adapted to many tasks.

Generative AI Concepts: Tokens, Embeddings, and Foundation Models (AIF-C01) — the lesson that teaches this.

What AIF-C01 domain 2 tests, topic by topic

The official exam guide breaks Fundamentals of GenAI into 3 topics. The question bank follows the same split, so a weak topic shows up as a cluster of misses you can go back and read.

Published AIF-C01 practice questions per topic in Fundamentals of GenAI
TopicWhat it coversQuestions
Explain the basic concepts of generative AI (GenAI)Exam guide task 2.1 (AIF-C01). Foundational GenAI concepts (tokens, chunking, embeddings, vectors, prompt engineering, transformer-based large language models [LLMs], foundation models [FMs], multi-modal models, diffusion models); potential GenAI use cases (image, video, and audio generation; summarization; AI assistants; translation; code generation; customer service agents; search; recommendation engines); the FM lifecycle (data selection, model selection, pre-training, fine-tuning, evaluation, deployment, feedback); the token-based pricing model and its effect on cost and performance for inference; the role of context engineering in FM applications; foundational agentic AI concepts (multi-agent system patterns, Model Context Protocol [MCP] and its role in connecting agents to external systems, multi-agent communication patterns, memory management, tool usage, workflow orchestration).34
Understand the capabilities and limitations of GenAI for solving business problemsExam guide task 2.2 (AIF-C01). Advantages of GenAI (adaptability, responsiveness, conversational capabilities, ability to generate content); disadvantages of GenAI solutions (hallucinations, interpretability, inaccuracy, nondeterminism); factors when selecting GenAI models (model types, performance requirements, capabilities, constraints, compliance, cost, latency, model complexity); business value and metrics for GenAI applications (cross-domain performance, ROI, efficiency, conversion rate, average revenue per user, accuracy, customer lifetime value).34
Describe AWS infrastructure and technologies for building GenAI applicationsExam guide task 2.3 (AIF-C01). AWS services and features to develop GenAI applications (Amazon Bedrock, Amazon SageMaker AI, SageMaker JumpStart, Amazon Quick, Kiro, Strands Agents, Amazon Bedrock AgentCore); advantages of AWS GenAI services (accessibility, lower barrier to entry, efficiency, cost-effectiveness, speed to market, ability to meet business objectives); benefits of AWS infrastructure for GenAI applications (security, compliance, responsibility, safety); cost tradeoffs of AWS GenAI services (responsiveness, availability, redundancy, performance, regional coverage, token-based pricing, provision throughput, custom models).34
Total102

Revise Fundamentals of GenAI before you drill it

Other AIF-C01 domains

Fundamentals of GenAI: your questions

Fundamentals of GenAI is domain 2 of the AIF-C01 exam guide and carries 24% of the scored content — the 2nd-heaviest of the 5 domains. On a 65-question paper that works out to roughly 16 questions, though AWS does not publish an exact per-domain count and individual exam forms vary.

Source

The domain weight and topic list on this page come from the official AIF-C01 exam guide.