- Introduction: Why AWS Services Dominate the AIF-C01
- AWS Bedrock: The Core of Generative AI on AWS
- Key Bedrock Features You Must Know
- Amazon SageMaker: ML Lifecycle from Build to Deploy
- Critical SageMaker Components for the Exam
- AWS Managed AI Services: The Ready-to-Use Layer
- Bedrock vs SageMaker vs AI Services: When to Use What
- Exam Strategy: How These Topics Appear in Questions
- Frequently Asked Questions
- If you are preparing for the AWS Certified AI Practitioner exam, there is one undeniable truth: you cannot pass AIF-C01 without a solid understanding of AWS...
- Amazon Bedrock is AWS's fully managed service for building generative AI applications using foundation models (FMs).
- The exam will test you on specific Bedrock capabilities.
- While Bedrock handles generative AI and foundation models, Amazon SageMaker is AWS's comprehensive platform for the full machine learning lifecycle - from data...
Introduction: Why AWS Services Dominate the AIF-C01
If you are preparing for the AWS Certified AI Practitioner exam, there is one undeniable truth: you cannot pass AIF-C01 without a solid understanding of AWS Bedrock, Amazon SageMaker, and the broader family of AWS AI services. Together, these tools form the backbone of roughly half the exam content, spanning Domains 2, 3, and 5 of the official exam guide.
These study notes are designed to give you a clear, structured understanding of each platform - what it does, how it differs from the others, and exactly how AWS tests your knowledge. Whether you are tackling an AWS AI Practitioner practice exam for the first time or doing a final review the night before your test, this guide covers the essential concepts you need to lock in.
The AIF-C01 is a foundational-level certification, but that does not mean it is easy. AWS expects you to understand not just what these services are, but when and why you would use each one in a real-world scenario. Let's break it all down.
AWS Bedrock: The Core of Generative AI on AWS
Amazon Bedrock is AWS's fully managed service for building generative AI applications using foundation models (FMs). It is the single most important service to understand for the AIF-C01 exam, appearing heavily across Domains 2 and 3.
At its core, Bedrock provides access to a curated selection of high-performing foundation models from leading AI companies - including Anthropic (Claude), Meta (Llama), Mistral AI, Cohere, and Amazon's own Titan and Nova model families - all through a unified API. You do not need to manage any underlying infrastructure. You simply select a model and call it.
What Makes Bedrock Different?
The key differentiator is that Bedrock is serverless and fully managed. Unlike SageMaker, which requires more hands-on ML engineering, Bedrock abstracts away all infrastructure concerns. Developers and architects can focus entirely on building applications rather than provisioning GPU instances or managing model deployments.
On the AIF-C01 exam, whenever a question describes a scenario where a company wants to use a pre-built large language model (LLM) without managing infrastructure, the answer almost always involves Amazon Bedrock. Memorize this pattern: no ML expertise required + pre-trained FM + managed service = Bedrock.
Bedrock also supports private customization. Organizations can fine-tune foundation models using their own proprietary data without that data being used to train the base model or shared with the model provider. This is a critical security and compliance feature that appears regularly in AIF-C01 exam questions.
Key Bedrock Features You Must Know
The exam will test you on specific Bedrock capabilities. Here are the ones that appear most frequently in AIF-C01 exam topics:
1. Model Customization: Fine-Tuning and Continued Pre-Training
Bedrock offers two methods for customizing foundation models with your own data:
- Fine-tuning: Adjusts model weights using labeled input-output examples. Best for teaching the model a specific task or style.
- Continued pre-training: Trains the model on large amounts of unlabeled domain-specific text. Best for deeply embedding industry knowledge (e.g., medical, legal, financial data).
2. Retrieval-Augmented Generation (RAG) with Knowledge Bases
Bedrock Knowledge Bases enable Retrieval-Augmented Generation (RAG), one of the most heavily tested concepts in the entire exam. RAG allows a foundation model to retrieve relevant information from an external knowledge source (like an S3 bucket full of company documents) before generating a response. This dramatically reduces hallucinations and keeps answers grounded in factual, up-to-date data.
3. Bedrock Agents
Bedrock Agents allow you to build autonomous AI agents that can plan and execute multi-step tasks by calling APIs, querying databases, and interacting with other AWS services. Agents are tested within the context of agentic AI workflows and orchestration in Domain 3.
4. Guardrails for Amazon Bedrock
Guardrails let you define safety policies for your generative AI application - filtering harmful content, blocking personally identifiable information (PII), and enforcing topic restrictions. This feature bridges Bedrock with Domain 4 (Responsible AI) and Domain 5 (Security and Governance).
Many candidates mix up fine-tuning with RAG. Fine-tuning changes the model's weights permanently using training data. RAG does NOT change the model - it retrieves external documents at inference time to supplement the model's response. The exam will absolutely test this distinction.
5. Model Evaluation
Bedrock includes built-in model evaluation tools that let you compare the performance of different foundation models on your specific tasks using custom datasets. This helps organizations choose the right model before committing to a production deployment.
Amazon SageMaker: ML Lifecycle from Build to Deploy
While Bedrock handles generative AI and foundation models, Amazon SageMaker is AWS's comprehensive platform for the full machine learning lifecycle - from data preparation and model training to deployment and monitoring. SageMaker is tested primarily in Domain 1 (Fundamentals of AI and ML) and to a lesser degree in Domain 3.
SageMaker is aimed at data scientists and ML engineers who need granular control over model development. It supports custom model training, hyperparameter tuning, distributed training across clusters, and real-time or batch inference endpoints.
Think of SageMaker as the engineer's toolkit and Bedrock as the developer's shortcut. SageMaker gives you maximum control and flexibility for custom ML models. Bedrock gives you immediate access to powerful pre-trained models with minimal setup. Both have a place in real-world AWS architectures - and the exam tests your ability to choose between them.
Critical SageMaker Components for the Exam
SageMaker is a vast service with dozens of features. For the AIF-C01 exam, focus on these key components:
The integrated development environment (IDE) for ML. Provides a unified interface for data exploration, model training, experimentation tracking, and deployment. Think of it as Jupyter notebooks on steroids, hosted by AWS.
A no-code visual interface that allows business analysts with no ML experience to build and deploy ML models. This is heavily relevant to the exam's focus on democratizing AI. If a question mentions a non-technical user building ML models, Canvas is likely the answer.
Automated Machine Learning (AutoML) that automatically explores different algorithms and hyperparameters to find the best model for your dataset. It provides full transparency - you can inspect the notebooks it generates. Key for exam questions about AutoML on AWS.
A CI/CD pipeline service for ML workflows, enabling MLOps best practices like automated retraining, model versioning, and deployment automation. Appears in exam scenarios about operationalizing ML models at scale.
Continuously monitors deployed models for data drift, model drift, and bias. Critical for responsible AI topics in Domain 4 and governance topics in Domain 5. If a question asks how to detect when a deployed model's performance degrades over time, Model Monitor is the answer.
Detects bias in datasets and model predictions, and provides explainability reports. This is one of the most important features for Domain 4 (Responsible AI). Know that Clarify can detect both pre-training bias (in the data) and post-training bias (in model outputs).
AWS Managed AI Services: The Ready-to-Use Layer
Beyond Bedrock and SageMaker, AWS offers a rich catalog of pre-built, task-specific AI services that require no ML knowledge whatsoever. These are called AWS AI Services, and they appear consistently across all five exam domains. For a complete breakdown of how these map to exam domains, see the AWS AI Practitioner (AIF-C01) Study Guide: Complete Exam Breakdown.
Here are the key managed AI services to know for the exam:
- Amazon Rekognition: Computer vision service for image and video analysis. Detects objects, faces, text, and inappropriate content. Used in media, security, and retail.
- Amazon Textract: Extracts text and structured data (forms, tables) from scanned documents. Goes beyond simple OCR by understanding document structure.
- Amazon Comprehend: Natural Language Processing (NLP) service for sentiment analysis, entity recognition, key phrase extraction, and topic modeling.
- Amazon Transcribe: Converts speech to text (STT). Supports custom vocabulary and speaker diarization.
- Amazon Polly: Converts text to speech (TTS). Supports multiple voices and languages, including neural voices.
- Amazon Translate: Neural machine translation service for real-time and batch text translation.
- Amazon Lex: Build conversational chatbots and voice interfaces using the same technology that powers Alexa. Understands natural language (NLU) and manages dialog flows.
- Amazon Kendra: Intelligent enterprise search powered by ML. Understands natural language queries and returns highly relevant answers from internal documents.
- Amazon Personalize: Real-time personalization and recommendation engine. Used by e-commerce and streaming platforms.
- Amazon Forecast: Time-series forecasting service using ML. Predicts future values like sales, inventory demand, or web traffic.
- Amazon Fraud Detector: Detects fraudulent activity online using ML models trained on AWS's fraud detection experience.
Don't try to memorize every API parameter for each service. Instead, focus on use case recognition: given a business problem, which AWS AI service solves it? The exam presents scenarios and asks you to select the most appropriate service. Practice this with an AWS AI Practitioner practice test to build pattern recognition quickly.
Bedrock vs SageMaker vs AI Services: When to Use What
One of the most common question types on the AIF-C01 exam presents a business scenario and asks which AWS service or approach is most appropriate. Understanding the positioning of each layer is essential.
| Criteria | AWS AI Services | Amazon Bedrock | Amazon SageMaker |
|---|---|---|---|
| Technical skill required | None (API calls) | Low (prompt engineering) | High (ML expertise) |
| Model customization | None | Fine-tuning / RAG | Full custom training |
| Infrastructure management | Fully managed | Fully managed | Managed but configurable |
| Best for | Specific tasks (vision, NLP, speech) | Generative AI applications | Custom ML model development |
| Primary exam domains | Domains 1, 3 | Domains 2, 3, 5 | Domains 1, 3, 4 |
| Example use case | Transcribe customer calls | Build an AI chatbot on company docs | Train a custom fraud detection model |
Understanding this three-layer model is a foundational concept not just for the exam, but for your career in cloud AI. If you want to explore how this certification fits into a broader learning journey, check out the AWS AI Certification Path: From AI Practitioner to ML Specialty to plan your next steps.
Many candidates assume SageMaker is always the answer for ML-related questions. This is wrong. If the scenario describes using a foundation model or generative AI without custom model training, Bedrock is the correct answer. If it describes a non-technical user building predictive models, SageMaker Canvas or Autopilot may be correct. Always read the scenario carefully for clues about technical complexity and use case type.
Exam Strategy: How These Topics Appear in Questions
Understanding the services is only half the battle. Knowing how AWS phrases questions about them is equally important. Here are the key patterns to watch for on your AIF-C01 practice test and on exam day:
Pattern 1: Scenario-Based Service Selection
These questions describe a business problem and ask you to select the most appropriate AWS service. Keywords to watch for:
- "Without managing infrastructure" → Bedrock or AWS AI Services
- "Existing internal documents" or "reduce hallucinations" → Bedrock Knowledge Bases (RAG)
- "Custom model" or "proprietary algorithm" → SageMaker
- "No ML expertise" + "predictions" → SageMaker Canvas or Autopilot
- "Monitor for bias" or "explain predictions" → SageMaker Clarify
- "Data drift" or "model degradation" → SageMaker Model Monitor
Pattern 2: Responsible AI and Governance
Domain 4 and 5 questions frequently reference Bedrock Guardrails, SageMaker Clarify, and SageMaker Model Monitor. Know that:
- Guardrails = content filtering and safety controls for generative AI
- Clarify = bias detection and explainability for ML models
- Model Monitor = ongoing drift detection in production
Pattern 3: Foundation Model Concepts
Expect multiple questions testing your understanding of prompt engineering, RAG, fine-tuning, and model evaluation - all in the context of Bedrock. For deeper preparation on these patterns, the AWS AI Practitioner Exam Tips: How to Pass on Your First Attempt guide offers proven strategies for tackling these question types.
Pattern 4: Cost and Efficiency Trade-offs
AWS loves to test whether you understand the cost and complexity trade-offs of different approaches. Fine-tuning a model is more expensive and time-consuming than using RAG. Training a custom SageMaker model is more expensive than calling a Bedrock API. These trade-offs appear regularly in scenario questions.
Spend at least 40% of your study time on Bedrock and its features (RAG, fine-tuning, Guardrails, Agents). Spend 25% on SageMaker key components (Canvas, Clarify, Model Monitor, Pipelines). Spend 20% on use-case recognition for AWS AI Services. Use the remaining 15% practicing scenario-based questions on a dedicated AWS AI Practitioner practice test platform to test your judgment across all these services.
For candidates who want to understand the difficulty level of these service-related questions in context, the article on How Hard Is the AWS AI Practitioner Exam? Pass Rate and Difficulty provides excellent perspective on what to expect.
Also, if you are still weighing whether to pursue this certification at all, the in-depth analysis at Is the AWS AI Practitioner Certification Worth It? Salary and Career Impact breaks down the ROI from multiple career angles.
Frequently Asked Questions
Amazon Bedrock is a fully managed service for accessing and customizing pre-built foundation models (LLMs) for generative AI applications - no ML expertise required. Amazon SageMaker is a comprehensive ML platform for data scientists and engineers to build, train, and deploy custom machine learning models with full control over the development process. On the AIF-C01 exam, Bedrock questions dominate Domains 2 and 3, while SageMaker questions are more common in Domain 1. The key distinction is: Bedrock for generative AI with pre-trained FMs, SageMaker for custom ML model development.
AWS does not publish exact question breakdowns by service, but given that Domain 2 (Generative AI Fundamentals, 24%) and Domain 3 (Applications of Foundation Models, 28%) together represent 52% of the exam, and Bedrock is the primary AWS service for both domains, you can expect a significant portion of questions to involve Bedrock concepts. Topics like RAG, fine-tuning, Guardrails, Bedrock Agents, and model evaluation are all high-frequency areas. Completing an AIF-C01 practice test will help you gauge the density of Bedrock questions.
The AIF-C01 is a foundational-level exam, and AWS states there are no prerequisites. However, familiarity with the AWS console and having used services like Bedrock or SageMaker Canvas - even through free-tier exploration - will significantly improve your ability to answer scenario-based questions. The exam tests conceptual understanding and service selection judgment more than hands-on technical skills, but real-world context helps tremendously. Reviewing an AWS AI Practitioner study guide alongside practical exploration is the ideal combination.
Based on the exam domain weightings and the topics covered in the official exam guide, the services appearing most frequently include: Amazon Bedrock (especially Knowledge Bases, Guardrails, and Agents), Amazon SageMaker (especially Clarify, Canvas, Model Monitor, and Autopilot), Amazon Rekognition, Amazon Comprehend, Amazon Lex, and Amazon Kendra. For each service, focus on its primary use case and the scenarios where it is preferred over alternatives. Use case recognition is the core skill being tested.
Both certifications serve different purposes. The AWS Cloud Practitioner (CLF-C02) validates foundational cloud knowledge across all AWS services. The AWS AI Practitioner (AIF-C01) specifically validates knowledge of AI, ML, and generative AI concepts and AWS implementations. If you work in or are moving toward AI, data, or ML roles, the AI Practitioner is highly relevant and signals specialized expertise. For a detailed comparison of both certifications, see the dedicated AWS AI Practitioner vs Cloud Practitioner: Which One First? guide.
Ready to Start Practicing?
Now that you have a solid understanding of AWS Bedrock, SageMaker, and the AWS AI Services ecosystem, it's time to test your knowledge. Our free AIF-C01 practice questions cover all five exam domains - including service selection scenarios, Bedrock feature questions, and responsible AI topics - so you can walk into your exam with confidence.
Start Free Practice Test →