AIF-C01 Cheat Sheet
AWS Certified AI Practitioner — Quick reference for the exam
CLI Commands
22aws bedrock list-foundation-modelsTOPList available foundation models
aws bedrock get-foundation-model --model-identifier <id>View details of a specific model
aws bedrock-runtime invoke-model --model-id <id> --body file://prompt.json out.jsonTOPInvoke a model for inference
aws bedrock create-knowledge-base --name <n> ...TOPCreate a Knowledge Base for RAG
aws bedrock create-agent --agent-name <n> --foundation-model <id>TOPCreate a Bedrock Agent
aws bedrock create-guardrail --name <n> ...TOPCreate a Guardrail (content/topic filters)
Comparisons
5| Service | Main Function | Input/Output Type |
|---|---|---|
| Amazon Comprehend | NLP: sentiment, entities, topics, PII | Text → insights |
| Amazon Rekognition | Image and video analysis, faces, objects | Image/video → labels |
| Amazon Textract | Extract text and data from documents/forms | Document → structured text |
| Amazon Transcribe | Speech to text (STT), includes diarization | Audio → text |
| Amazon Polly | Text to speech (TTS), neural voices | Text → audio |
| Amazon Lex | Conversational chatbots (NLU + ASR) | Voice/text → intents |
| Amazon Kendra | Enterprise semantic search | Question → relevant document |
| Amazon Bedrock | API access to foundation models (GenAI) | Prompt → generated text/image |
| Amazon SageMaker | Build, train, and deploy your own ML models | Data → trained model |
Quick Reference
37ML/AI Fundamentals (D1)
AI is the general field. ML learns from data. Deep Learning uses neural networks
Trains on labeled data (classification, regression)
Finds patterns without labels (clustering, anomalies)
Learns by trial and error, maximizing a reward
The model memorizes the training set and fails to generalize to new data
Detects bias and explains model predictions
Generative AI Concepts (D2)
Large-scale pre-trained model, adaptable to multiple tasks
Smallest unit of text processed by the model
0 = deterministic/precise. High = more creative/random
Control how many next-word options are considered
Numeric vector representation of text for semantic search
The model generates false information with an appearance of certainty
Maximum number of tokens the model can process at once
Agentic AI and MCP (D2/D3)
LLM + ability to reason, plan, and execute actions/tools
Model Context Protocol — open standard for connecting models to tools/data
AWS managed runtime for deploying production agents
Several specialized agents collaborating on a complex task
The agent decides which tool to invoke and in what order
Persistent context across an agent's turns/sessions
RAG and Fine-tuning (D3)
Retrieval-Augmented Generation — grounds responses in your own data
Stores embeddings for similarity search (OpenSearch, Aurora, Kendra)
Splitting documents into fragments before generating embeddings
Retraining the model with your own data to specialize it
Training with unlabeled domain data before fine-tuning
A small model ("student") learns from a large one ("teacher")
Responsible AI (D4)
Systematically unfair outcomes toward a group
Document describing a model's capabilities, limits, and intended use
Configurable filters for topics, harmful content, and PII
Human review of high-impact decisions (Amazon A2I)
Clearly explaining how and with what data the model was trained
Security and Governance (D5)
AWS manages the model/infrastructure. You manage prompts, data, IAM
AWS's 5 scopes for classifying the risk of GenAI apps
Audits configuration compliance of AI resources
Collects continuous evidence for compliance audits
Access to AWS compliance reports and certifications
Connects to AI services without exposing traffic to the public internet
Where data is stored/processed according to regulatory requirements
Ready to practice?
Put what you learned to the test with original AIF-C01 exam questions.
Start AIF-C01 mock examUnlimited mock exams with the Pro plan