FM Design, RAG and Agents
This topic maps foundation model selection, inference parameters, Retrieval Augmented Generation, vector databases, customization tradeoffs, and AI agent use cases.
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 model selection depends on modality, cost, latency, language, input and output length, capability, customization, and complexity.
Exam cue: Use RAG when an application needs current or proprietary knowledge without retraining the model.
Concept 2
Retrieval Augmented Generation grounds responses by retrieving relevant context from external knowledge sources.
Exam cue: Use temperature and length parameters when output variation or response size matters.
Concept 3
AI agents combine model reasoning with tools, instructions, memory, and workflow orchestration to complete tasks.
Exam cue: Use agents when the system must plan and invoke tools across steps.
Risk pitfalls and guardrails
Fine-tuning a model just to add searchable company knowledge.
Guardrail: Avoid choosing GenAI because it sounds modern, trusting fluent output without validation, or ignoring privacy, cost, and governance requirements.
Ignoring input and output length limits.
Guardrail: Avoid choosing GenAI because it sounds modern, trusting fluent output without validation, or ignoring privacy, cost, and governance requirements.
Treating a vector database as a replacement for governance or source quality.
Guardrail: Avoid choosing GenAI because it sounds modern, trusting fluent output without validation, or ignoring privacy, cost, and governance requirements.
Memory anchors
RAG
Retrieval Augmented Generation retrieves relevant context and supplies it to a model before generation.
Knowledge Base
A knowledge base organizes external content for retrieval in a GenAI application.
Temperature
Temperature controls how varied or deterministic generated responses tend to be.
Input Length
Input length is the amount of prompt and context a model can accept.
Output Length
Output length controls or limits the amount of generated response.
Prompt Caching
Prompt caching can reduce latency or cost when repeated context is reused.
AI Agent
An AI agent uses a model plus tools or workflow steps to pursue a user goal.
Model Distillation
Model distillation trains a smaller model to approximate behavior from a larger model.
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
A company needs a model to create captions for product images. Which model-selection criterion should be treated as mandatory?
A creative-writing application produces repetitive responses. Which inference setting can be increased moderately to encourage more varied token choices?
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.”