GenAI Foundational Concepts
This topic covers tokens, chunking, embeddings, vectors, prompt engineering, foundation models, multimodal models, diffusion models, context engineering, and agentic AI concepts.
How to study for AWS Certified AI Practitioner
Treat each question as a business and governance decision: identify the AI pattern, choose the right AWS capability, then add responsible AI, cost, security, and evaluation controls.
Core concepts
Concept 1
Foundation models are large models that can be adapted to many tasks through prompts, retrieval, fine-tuning, or other customization methods.
Exam cue: Identify whether the question is asking about model input units, semantic search, generation, or agent workflow.
Concept 2
Tokens, chunks, embeddings, and vector search are core mechanics in many GenAI and retrieval workflows.
Exam cue: Use embeddings and vectors when the scenario needs semantic similarity.
Concept 3
Agentic AI extends model use with tool calls, memory, workflow orchestration, and communication patterns.
Exam cue: Use context engineering when better inputs and external context are the focus.
Risk pitfalls and guardrails
Treating tokens as words in every language and model.
Guardrail: Avoid choosing GenAI because it sounds modern, trusting fluent output without validation, or ignoring privacy, cost, and governance requirements.
Expecting embeddings to store source documents without a vector store or retrieval step.
Guardrail: Avoid choosing GenAI because it sounds modern, trusting fluent output without validation, or ignoring privacy, cost, and governance requirements.
Calling simple chat completion an autonomous agent.
Guardrail: Avoid choosing GenAI because it sounds modern, trusting fluent output without validation, or ignoring privacy, cost, and governance requirements.
Memory anchors
Token
A token is a unit of model input or output used for processing and pricing.
Chunking
Chunking splits source content into smaller units for retrieval or processing.
Embedding
An embedding represents meaning as numeric values that can be compared.
Vector Store
A vector store indexes embeddings so semantically similar content can be retrieved.
Foundation Model
A foundation model is a large model trained broadly and adapted for many tasks.
Multimodal Model
A multimodal model can work across input or output types such as text, image, audio, or video.
Context Engineering
Context engineering designs the information supplied to a model so it can respond usefully.
Tool Use
Tool use lets an AI application call external systems to retrieve information or take action.
Checkpoint rule
Do the check-up only after you can summarize each concept in one sentence and identify one dangerous pitfall from memory.
Knowledge Check (after reading)
Short check-up to confirm understanding of this module.
Check-up Questions
In a foundation model application, what is a token?
Why is it inaccurate to assume that one token always equals one English word?
Answer all questions to submit.
Next step personalized recommendations
Continue learning
Move forward only after this module is stable.
What is Pass Harbor?
Completely free exam prep for 317 U.S. exams.
- Practice questions
- Flashcards
- Study guides
- Mock exams
- No registration
- No paywall
- Start instantly
“No more expensive exam prep. Quality study tools should be accessible to everyone.”