AWS AI Practitioner (AIF-C01) Cheat Sheet 2026: Complete Tutorial & Exam Guide

Introduction to the AWS Certified AI Practitioner Exam

Exam overview

• Exam duration: 90 minutes
• Exam format: 65 questions

Content domains and weightings:

• Domain 1: Fundamentals of AI and ML (20% of scored content)
• Domain 2: Fundamentals of Generative AI (24% of scored content)
• Domain 3: Applications of Foundation Models (28% of scored content)
• Domain 4: Guidelines for Responsible AI (14% of scored content)
• Domain 5: Security, Compliance, and Governance for AI Solutions (14% of scored content)

Start Practicing Now

AWS Certified AI Practitioner (AIF-C01) Practice Tests

Domain 1: Fundamentals of AI and ML (20%)

Task Statement 1.1: Explain basic AI concepts and terminologies.

What is Artificial Intelligence?
Artificial Intelligence (AI) is a field of computer science focused on developing systems that can exhibit intelligent behavior, such as reasoning, learning, and autonomous action. AI functions by combining large amounts of data with intelligent algorithms. AWS offers pre-built AI services and customizable infrastructure options.

How does AI process information?

• Data Collection: vast amounts of data
• Algorithm Selection: ML, Deep Learning, NLP
• Training: algorithm learns from data
• Testing: evaluate performance on new data
• Deployment: real-world application

Key components of AI application architecture: Data layer, Model layer, Application layer.

Applications of AI: Chatbots, intelligent document processing, APM, predictive maintenance, medical research, business analytics.

Limitations of AI on AWS: Data quality & bias, computational cost, ethical concerns, need for human oversight.

Basic AI Terminologies

Machine Learning (ML): Algorithms that improve automatically through experience and data.
Deep Learning: Subset of ML using multilayered neural networks.
Large Language Model (LLM): Deep learning models trained on extensive text using transformer architecture.
Responsible AI: Fairness, transparency, ethical practices.
Neural networks: Interconnected computational units inspired by human brain.
Natural Language Processing (NLP): Extracts meaning from text.
Computer vision: Interprets visual information from images/videos.
Speech recognition: Deciphers human speech into text.
Generative AI: Produces original content (images, text, audio) based on prompts.

Differences between AI, ML, Deep Learning & Gen AI

• AI: broad field, rule-based, robotics, NLP (Siri)
• ML: subset, learns from data (spam filters)
• Deep Learning: neural networks with layers (facial recognition)
• Generative AI: GANs, VAEs, LLMs (ChatGPT, DALL-E)

ML Types: Supervised (labeled data), Unsupervised (unlabeled, clustering), Reinforcement Learning (reward system).
ML Terms: Features, Training Data, Overfitting, Underfitting.

Task Statement 1.2: Identify practical use cases for AI.

AWS Managed AI Services: SageMaker, Transcribe, Translate, Comprehend, Lex, Polly, Rekognition, Textract, Personalize, Forecast, Kendra, IoT Greengrass ML Inference, Neptune ML, Amazon Mechanical Turk (MTurk), Amazon Augmented AI (A2I), AWS DeepRacer.

Task Statement 1.3: Describe the ML development lifecycle (ML Pipeline with AWS services).

• Data Collection: Amazon S3, AWS Glue, Kinesis, RDS
• Data Pre-processing / EDA: SageMaker Data Wrangler, AWS Glue, SageMaker Studio, Athena, SageMaker Processing
• Feature Engineering: SageMaker Feature Store, Data Wrangler
• Model Training: SageMaker, Deep Learning AMIs, Spot Instances
• Hyperparameter Tuning: SageMaker Automatic Model Tuning
• Evaluation: SageMaker Debugger, Experiments, Model Monitor
• Deployment: SageMaker Endpoints, API Gateway, Lambda, EKS
• Monitoring: SageMaker Model Monitor, CloudWatch, SageMaker Pipelines, X-Ray

Fundamentals of ML Operations (MLOps)

• Experimentation (SageMaker notebooks, Experiments)
• Repeatable processes (SageMaker Pipelines, IaC)
• Scalable systems (EC2, distributed training)
• Technical debt mgmt (version control, Model Registry)
• Production readiness (CI/CD, IAM, KMS)
• Model monitoring & re-training (Model Monitor, Step Functions, drift detection)
• Multi-model endpoints, collaboration (SageMaker Studio), governance (Clarify)

Amazon SageMaker – Comprehensive Platform

Algorithm selection examples: Logistic Regression, XGBoost, Linear Learner, Faster R-CNN, Random Cut Forest, K-Means, LDA, Factorization Machines.

Features: Feature Store, Data Wrangler, Geospatial ML, Notebooks, JumpStart, Studio Lab, Model Training, Experiments, HyperPod, Pipelines, MLOps, Canvas, Ground Truth, Clarify, Role Manager.

SageMaker Canvas vs Studio: Canvas (no-code, business analysts) – Studio (IDE, advanced coding).

Domain 2: Fundamentals of Generative AI (24%)

What is Generative AI?

Generative AI creates original content (text, images, audio). Uses foundation models (FMs) trained on massive datasets. LLMs are a class of FMs.

Benefits: accelerates research, enhances customer experience, optimizes processes, boosts productivity.

Generative AI Models: Diffusion models, GANs (generator/discriminator), Variational Autoencoders (VAEs), Transformer-based models (self-attention).

Tools for building gen AI apps: Amazon Bedrock, SageMaker, AWS Trainium, Inferentia, EC2 P5, EC2 UltraClusters.

Applications: Amazon Q, PartyRock, AWS HealthScribe. Use cases: chatbots, code generation, personalization, conversational analytics.

Limitations: security (data privacy), creativity constraints, cost, explainability (“black boxes”).

Amazon SageMaker JumpStart

ML hub with foundation models from AI21, Hugging Face, Meta, Mistral, Stability AI. Built-in algorithms and prebuilt solutions. Benefits: publicly available FMs, no-code deployment, support for LLaMA 2, Stable Diffusion, text classification, image generation.

Amazon Bedrock

Serverless API for foundation models (AI21, Anthropic, Cohere, Meta, Mistral, Stability, Amazon). Unified API, customization via fine-tuning, RAG, Agents. Model states: Active, Legacy, EOL. Use cases: text generation, virtual assistants, search, summarization, image generation.

Bedrock Agents: access company data, orchestrate tasks, dynamic code execution, memory across interactions.
Bedrock Guardrails: customizable safeguards against harmful content, hallucinations, prompt attacks. Contextual grounding checks.

PartyRock: playground within Bedrock for chat, text, image experimentation.

Amazon Q

Amazon Q Business: AI assistant for enterprise data (answers, summaries, tasks).
Amazon Q Developer: coding companion for Java, Python, JS, C#, etc., real-time suggestions, RAG customization. Integrates with QuickSight, Connect, AWS Supply Chain.

GenAI Security Scoping Matrix

• Scope 1: Public consumer apps (PartyRock, ChatGPT) – no ownership
• Scope 2: Enterprise apps with AI features (Amazon Q)
• Scope 3: Pre-trained models (Bedrock base models)
• Scope 4: Fine-tuned models (Bedrock custom models, JumpStart)
• Scope 5: Self-trained models from scratch (SageMaker)

Security disciplines: governance & compliance, legal & privacy, risk management, controls, resilience.

Domain 3: Applications of Foundation Models (28%)

Understanding Foundation Models (FMs)

Large-scale deep neural networks trained on extensive datasets. Features: adaptability, general-purpose nature. Applications: language processing, visual comprehension, code generation, decision support. Examples: BERT, GPT (OpenAI), Amazon Titan. Challenges: high resource demands, integration complexity, reliability issues.

Design considerations for FM applications

• Model selection (text vs image vs code)
• Cost management (use Bedrock, JumpStart)
• Prompt design & engineering
• Fine-tuning with domain data
• Performance & scalability (auto-scaling, serverless)
• Ethical & responsible AI (bias, safeguards)
• Data privacy & security (encryption, access controls)
• Model monitoring (SageMaker Model Monitor)
• Integration (API Gateway, Lambda)

Prompt Engineering Techniques

• Chain-of-thought: divide complex questions into logical steps
• Tree-of-thought: generate multiple next steps, tree search
• Maleutic prompting: explain parts, eliminate inconsistencies
• Complexity-based prompting: multiple chain-of-thought rollouts
• Generated knowledge prompting: generate facts before answering
• Least-to-most prompting: solve subproblems sequentially
• Self-refine prompting: solve, evaluate, improve iteratively
• Directional-stimulus prompting: hints or keywords to guide model
• Few-shot learning: provide examples; Zero-shot: clear instructions only

Retrieval Augmented Generation (RAG)

Enhances LLMs by referencing authoritative knowledge base beyond training data. Steps: create external data → retrieve relevant info (vector DB) → augment LLM prompt → update external data asynchronously. Benefits: cost-effective, current info, user trust, developer control.

RLHF – Reinforcement Learning from Human Feedback

Enhances model performance using human feedback. Process: data collection → supervised fine-tuning (using RAG) → build reward model (human preferences) → optimize language model with reward model. Applications: image generation, music, voice assistant tone. AWS service: SageMaker Ground Truth.

Domain 4: Guidelines for Responsible AI (14%)

Responsible AI principles

• Fairness (diverse datasets, bias monitoring)
• Transparency (explainable decisions)
• Accountability (audit trails)
• Privacy and security (encryption, GDPR/HIPAA)
• Ethical use (prevent harmful applications)
• Continuous monitoring (drift, feedback)
• Accessibility (inclusive design)

Guardrails: mechanisms for safe, reliable AI behavior.
Bias: error from simplistic assumptions → underfitting. Variance: over-sensitivity to training data → overfitting.

Tools for detection: SageMaker Clarify (bias detection, feature importance), SageMaker Model Monitor (drift, anomalies), Amazon Bedrock (trustworthy FMs), custom fine-tuning, human-in-the-loop (Amazon A2I).

Domain 5: Security, Compliance and Governance for AI Solutions (14%)

Key concepts

• Security: confidentiality, integrity, availability
• Compliance: adherence to regulations (healthcare, finance, legal)
• Governance: define and enforce responsible AI practices
• Safety: beneficial algorithms
• Veracity & robustness: reliable under unexpected situations

AWS Security, Compliance & Governance Services

• IAM Users / Groups / Policies / Roles
• Amazon Macie (sensitive data in S3)
• AWS Config (track changes & compliance)
• Amazon Inspector (vulnerabilities in EC2, Lambda)
• AWS CloudTrail (API call tracking)
• AWS Artifact (compliance reports: PCI, ISO)
• AWS Trusted Advisor (insights, best practices)
• AWS PrivateLink (private connectivity)

Last Touch: Key Service Summaries

• SageMaker: build, train, deploy ML models; full ML workflow.
• SageMaker Studio: visual IDE for ML.
• SageMaker Debugger: periodic model state, rules for unwanted conditions.
• SageMaker Model Monitor: alerts on quality deviations, anomalies, drift via CloudWatch.
• SageMaker Clarify + Model Monitor: bias detection and alerts.
• SageMaker Clarify: explainability, feature contributions.
• SageMaker JumpStart: 150+ open source models (NLP, object detection, image classification).
• SageMaker Data Wrangler: import, transform, analyze data.
• SageMaker Feature Store: share/discover features; online (low latency) or offline (batch).
• SageMaker Canvas: no-code ML for business analysts (upload CSV, predict).
• Amazon Bedrock: API for FMs; serverless, supports RAG, agents, third-party model billing.
• Amazon Q Developer: AI coding companion (Java, Python, JS, C#, etc.) real-time suggestions, RAG customization.
• Amazon Q Business: enterprise assistant using RAG; answers, summaries, content generation.

Start Practicing Now

AWS Certified AI Practitioner (AIF-C01) Practice Tests

Recommended Articles

• AWS AI Practitioner Certification Guide 2026
• Generative AI Certification Preparation Guide
• AWS vs Azure vs Google Cloud Certifications 2026