AWS AIF-C01 Cheat Sheet: AI Practitioner
Review a compact AWS Certified AI Practitioner (AIF-C01) cheat sheet for AI and machine learning fundamentals, generative AI, foundation-model applications, responsible AI, and AWS security and governance before using IT Mastery practice.
Use this cheat sheet as a quick vocabulary and decision-pattern review before trying the free diagnostic or focused AIF-C01 topic questions. It keeps common AI, generative AI, and AWS governance distinctions clear before practice.
Snapshot
| Item | Review cue |
|---|---|
| Exam route | AWS Certified AI Practitioner |
| Exam code | AIF-C01 |
| Items | 65 total |
| Time | 90 minutes |
| Practice option | Live IT Mastery practice available |
| Best use | Build vocabulary, then practice scenario judgment across AI value, model choice, responsible AI, and governance |
Domain checklist
| Domain | Weight | What to know | Common trap |
|---|---|---|---|
| Fundamentals of AI and ML | 20% | supervised vs unsupervised learning, training vs inference, prediction vs classification, data quality | treating every AI task as generative AI |
| Fundamentals of Generative AI | 24% | prompts, tokens, foundation models, embeddings, retrieval augmented generation, hallucination risk | assuming a larger model is always better |
| Applications of Foundation Models | 28% | chat, summarization, search, content generation, agents, business fit | choosing GenAI when a simpler AWS service fits |
| Responsible AI | 14% | fairness, transparency, privacy, explainability, human review, bias | ignoring output risk because the prompt looks simple |
| Security, Compliance, and Governance | 14% | IAM, data protection, encryption, logging, policy, guardrails | focusing only on model quality and forgetting data boundaries |
AI practitioner decision map
Use the map to keep AIF-C01 choices grounded. The exam often gives a business need first, then expects you to choose an AI pattern, AWS service fit, and governance control instead of jumping straight to a model name.
flowchart LR
Need["Business need"] --> Pattern["AI / ML / GenAI pattern"]
Pattern --> Fit["AWS service and model fit"]
Fit --> Control["Responsible AI and security controls"]
Control --> Practice["Review missed decision rule"]
Must-know distinctions
| Distinction | Exam reflex |
|---|---|
| AI vs ML vs GenAI | AI is the broad field, ML learns from data, GenAI produces new content. |
| Training vs inference | Training builds or adapts the model. Inference uses the model to generate predictions or outputs. |
| Prompt engineering vs fine-tuning | Prompting changes instructions. Fine-tuning changes model behavior through additional training. |
| RAG vs fine-tuning | Retrieval augmented generation grounds answers in retrieved context. Fine-tuning adapts style or task behavior. |
| Embeddings vs generated text | Embeddings represent meaning for search and similarity. Generated text is the output users read. |
| Accuracy vs hallucination control | Correctness needs grounding, evaluation, constraints, and sometimes human review. |
| Bias vs privacy | Bias is unfair or distorted behavior. Privacy is protection of sensitive data. Both can matter in one scenario. |
Snippets to recognize
AIF-C01 is not a coding exam, but public sample scenarios may show short prompt, policy, or retrieval patterns. Recognize the risk or service decision behind the snippet.
Task: Summarize customer-specific support history.
Need: current private context
Better pattern: retrieve approved records first, then ask the foundation model to summarize only that context.
Trap: asking the model to invent missing account history from general knowledge.
{
"prompt_control": "Do not include personal data that is not present in the retrieved context.",
"review_required": true,
"log_sensitive_fields": false
}
High-yield checklist
- Identify the business problem before picking an AI service.
- Ask whether the scenario needs prediction, classification, recommendation, search, summarization, or generation.
- Prefer managed AWS AI services when the task is common and does not need a custom model.
- Use foundation models when language, reasoning, summarization, or content generation is central.
- Use RAG when answers must use current or private knowledge.
- Protect sensitive data in prompts, logs, datasets, vector stores, and outputs.
- Use guardrails, human review, or workflow controls when output harm is possible.
- Evaluate AI systems with representative prompts and expected behavior, not just one successful demo.
- Balance model quality, latency, cost, safety, and governance.
Practice strategy
For each missed AIF-C01 question, label the miss as vocabulary, service fit, responsible AI, or governance. If misses cluster in vocabulary, review the distinctions above. If misses cluster in service fit, move to topic drills before another timed set.