Training course

This is a free, self-paced reading course for AIF-C01: AWS Certified AI Practitioner, the foundational credential for people who make decisions about AI, machine learning, and generative AI on AWS without necessarily building the models themselves. You will not write code or tune hyperparameters for this exam, and the course does not teach you to. It teaches the vocabulary and the judgment the exam actually measures: when generative AI fits a business problem and when it does not, whether prompt engineering, RAG, or fine-tuning is the right answer, which managed AWS service solves which task, and how bias, transparency, security, and governance obligations constrain what you can ship.

The course is organized as one module per official exam domain, in the order AWS publishes them, and each module carries the domain’s published weight. That matters more here than on most foundational exams, because AIF-C01 is not evenly spread: generative AI and the applications of foundation models together account for more than half the exam, so they account for more than half the course. Every lesson ends with its key terms and suggested further reading, so it works as a reference on the second pass as well as a syllabus on the first.

AIF-C01 Foundational level 5 modules Domain-weighted Self-paced Free account

What the course covers

Fundamentals of AI and ML

Module 1 · 20%

The core vocabulary: AI versus ML versus deep learning versus generative and agentic AI; supervised, unsupervised, and reinforcement learning; classification, regression, clustering, and anomaly detection; overfitting, bias, and evaluation metrics such as precision, recall, and F1; inferencing from batch to real time; the ML development lifecycle and MLOps; and the question the exam keeps asking, whether AI is the right tool at all. Ends with service-to-use-case mapping across SageMaker, Comprehend, Transcribe, Textract, Rekognition, Personalize, Forecast, and Lex.

Fundamentals of Generative AI

Module 2 · 24%

How foundation models work and what they are good for: tokens, embeddings, context windows, transformers, and diffusion models; multimodality; inference parameters including temperature, top-p, and top-k; hallucination, nondeterminism, and the limits of interpretability; and the foundation model lifecycle. Covers the AWS generative-AI stack (Amazon Bedrock, Amazon Q Business and Q Developer, SageMaker JumpStart, and PartyRock) with the on-demand versus provisioned-throughput cost trade-off you are expected to reason about.

Applications of Foundation Models

Module 3 · 28%

The heaviest and most technical module. Model-selection criteria; prompt engineering from zero-shot and few-shot to chain-of-thought and negative prompting; prompt attacks, injection, jailbreaking, and leaking; Retrieval Augmented Generation with Bedrock Knowledge Bases and vector stores such as OpenSearch and Aurora pgvector; the customization spectrum running from prompting through RAG and fine-tuning to continued pre-training; Bedrock Agents, Guardrails, and Prompt Management; and evaluation with ROUGE, BLEU, BERTScore, human review, and business-alignment metrics.

Guidelines for Responsible AI

Module 4 · 14%

Fairness, the sources of bias and what dataset representativeness means in practice, transparency and explainability, veracity, safety, and human oversight, then the tooling that turns those principles into something auditable: SageMaker Clarify for bias detection and explainability, Amazon A2I for human review, SageMaker Model Cards for documentation, and AWS AI Service Cards for provider transparency. The exam asks these as scenarios about interpretable-versus-accurate model trade-offs and the legal exposure created by hallucinated content.

Security, Compliance, and Governance for AI Solutions

Module 5 · 14%

Securing AI systems and governing the data behind them: IAM least privilege, KMS encryption, PrivateLink and VPC endpoints for private model access, Macie for sensitive-data discovery, CloudTrail and Bedrock model invocation logging for auditability, and where the shared responsibility model draws its line. Governance covers data lineage and cataloging, retention, vendor due diligence, AWS Artifact, AWS Config, Audit Manager, and the external frameworks the exam names, the NIST AI RMF and ISO/IEC 42001.

How to use it

Read the modules in order rather than skipping to the heavy ones. AIF-C01 is a vocabulary exam before it is anything else, and modules 2 and 3 assume the definitions module 1 sets up, you cannot reason about whether RAG or fine-tuning fits a scenario until inference, embeddings, and context windows are settled. Once a module is read, test it before moving on: the difference between recognizing the phrase “chain-of-thought prompting” and picking it correctly out of four plausible options in a business scenario is the difference this exam is built to find. The course modules themselves open once you are signed in to a free Certifym account.

For exam logistics (question count, timing, the scaled cut score, pricing, and how long the credential stays valid) and for the domain-weighted practice exam that pairs with these modules, see the AIF-C01 certification guide. If you want depth rather than breadth after this course, the MLA-C01 training course is the associate-level step that takes the same subject matter into building and operating production ML systems.

AIF-C01: AWS Certified AI Practitioner

One module per AIF-C01 exam domain, weighted the way AWS weights them.

A self-paced reading course for AIF-C01: AWS Certified AI Practitioner, the foundational credential for people who make decisions about artificial intelligence, machine learning, and generative AI on AWS without necessarily building the models themselves. The course is organized as one module per official exam domain, in the order AWS publishes them, and each module carries the domain's published weight, so the time you spend maps to the way the exam distributes its questions.

You work through the core AI and machine learning vocabulary, the mechanics of foundation models and generative AI, the applications of those models — prompt engineering, Retrieval Augmented Generation, fine-tuning, agents, and evaluation — and then the two domains that decide whether a working prototype can actually ship: responsible AI, and security, compliance, and governance. Every lesson ends with its key terms and further reading, so the course works as a reference on the second pass as well as a syllabus on the first.

AIF-C01 does not ask you to write code or tune hyperparameters, and this course does not teach you to. It teaches judgment: when generative AI fits a business problem and when it does not, whether prompt engineering, RAG, or fine-tuning is the right answer to a given constraint, which managed AWS service solves which task, and how bias, transparency, and regulatory obligations shape what you are allowed to deploy. AWS revises its services and its exam guide frequently, so verify current service behavior, quotas, and pricing in the AWS documentation as you study. Certifym is an independent study publisher and is not affiliated with, endorsed by, or sponsored by Amazon Web Services, Inc.

Module 1: Fundamentals of AI and ML 20% of exam

The vocabulary domain, and the one every other module leans on. You start by separating the terms that get used interchangeably in conversation and never interchangeably on the exam — artificial intelligence, machine learning, deep learning, generative AI, and agentic AI — then work through the three learning paradigms, the shape of the data each one needs, and the problem types that map to them.The second half is practitioner judgment: the machine learning development lifecycle from business framing to monitoring, why models overfit and how you detect it, which evaluation metric answers which question, the inference options and their cost profiles, what MLOps adds once a model is in production, and finally the service-to-use-case mapping across the AWS AI and ML portfolio — including the cases where the correct answer is that machine learning is the wrong tool.

1.1 AI, machine learning, deep learning, generative AI, and agentic AI 5 min Free preview
1.2 Supervised, unsupervised, and reinforcement learning 5 min 🔒
1.3 Data for machine learning: structure, labeling, and quality 5 min 🔒
1.4 Recognizing machine learning problem types 5 min 🔒
1.5 The machine learning development lifecycle 5 min 🔒
1.6 Overfitting, underfitting, and the bias-variance trade-off 5 min 🔒
1.7 Evaluating models: technical metrics and business metrics 5 min 🔒
1.8 Inference options: batch, asynchronous, real time, and edge 5 min 🔒
1.9 MLOps: operating models in production 5 min 🔒
1.10 Choosing AWS AI and ML services 5 min 🔒

Module 2: Fundamentals of Generative AI 24% of exam

How foundation models actually work, and what follows from the way they work. The module opens with what makes a foundation model different from a task-specific model, then builds the mechanics from the bottom up: tokens, embeddings and vector representations, the transformer and diffusion architectures, multimodality, and the context window that bounds every interaction. From there you learn the inference parameters — temperature, top-p, top-k, and length controls — that shape output at request time.The second half covers the consequences and the platform. Hallucination, nondeterminism, and the limits of interpretability are properties of the technology rather than defects to be patched, and they determine which use cases are viable. You then work through the foundation model lifecycle and the AWS generative AI stack — Amazon Bedrock, Amazon Q Business and Q Developer, SageMaker JumpStart, and PartyRock — closing with the cost and throughput trade-offs the exam expects you to reason about.

2.1 Foundation models and what makes them different 5 min 🔒
2.2 Tokens, embeddings, and vector representations 5 min 🔒
2.3 Transformers, diffusion models, and multimodality 5 min 🔒
2.4 The context window and what fits inside it 5 min 🔒
2.5 Inference parameters: temperature, top-p, top-k, and length 5 min 🔒
2.6 Hallucination, nondeterminism, and the limits of interpretability 5 min 🔒
2.7 The foundation model lifecycle 5 min 🔒
2.8 Generative AI use cases, advantages, and limitations 5 min 🔒
2.9 Amazon Bedrock: the managed foundation model platform 5 min 🔒
2.10 Amazon Q Business and Amazon Q Developer 5 min 🔒
2.11 SageMaker JumpStart, SageMaker AI, and PartyRock 5 min 🔒
2.12 Cost and throughput trade-offs in generative AI 5 min 🔒

Module 3: Applications of Foundation Models 28% of exam

The heaviest and most technical domain, and the one where the exam stops asking what things are and starts asking which one you would choose. It opens with model selection criteria, then works through prompt engineering — the anatomy of a prompt, the named techniques from zero-shot to chain-of-thought, and the attacks that target prompts: injection, jailbreaking, and leaking.The center of the module is Retrieval Augmented Generation: what it does, the vector store options on AWS, and how Amazon Bedrock Knowledge Bases packages the whole pattern. From there you place RAG on the customization spectrum alongside prompt engineering, fine-tuning, and continued pre-training, and learn the criteria that decide between them. The module closes with agents and tool use, Guardrails, Prompt Management and Flows, and evaluation — ROUGE, BLEU, BERTScore, human review, and the business-alignment metrics that decide whether any of it worked.

3.1 Selecting a foundation model 5 min 🔒
3.2 Prompt engineering: the anatomy of a prompt 5 min 🔒
3.3 Prompting techniques: zero-shot, few-shot, and chain-of-thought 5 min 🔒
3.4 Prompt attacks: injection, jailbreaking, and leaking 5 min 🔒
3.5 Retrieval Augmented Generation 5 min Free preview
3.6 Vector stores and retrieval infrastructure on AWS 5 min 🔒
3.7 Amazon Bedrock Knowledge Bases 5 min 🔒
3.8 The customization spectrum 5 min 🔒
3.9 Fine-tuning in practice 5 min 🔒
3.10 Agents, tool use, and multi-step workflows 5 min 🔒
3.11 Guardrails and output safety controls 5 min 🔒
3.12 Prompt Management, Flows, and Data Automation 5 min 🔒
3.13 Evaluating generative model outputs 5 min 🔒
3.14 Business-alignment metrics and production monitoring 5 min 🔒

Module 4: Guidelines for Responsible AI 14% of exam

The domain that decides whether a working system is allowed to ship. It opens with the dimensions AWS uses to frame responsible AI — fairness, explainability, privacy and security, safety, controllability, veracity and robustness, governance, and transparency — then works through where bias actually comes from, how dataset representativeness fails in practice, and how SageMaker Clarify measures both bias and feature attribution.The rest of the module covers the mechanisms that turn principles into something auditable: the interpretability-versus-accuracy trade-off and when an unexplainable model is simply not permitted, human oversight through Amazon Augmented AI, documentation through SageMaker Model Cards and AWS AI Service Cards, and the veracity, safety, and legal exposure created by generated content. Expect scenario questions rather than definitions.

4.1 The dimensions of responsible AI 5 min 🔒
4.2 Where bias comes from 5 min 🔒
4.3 Measuring bias and explaining predictions with SageMaker Clarify 5 min 🔒
4.4 Transparency, explainability, and the interpretability trade-off 5 min 🔒
4.5 Human oversight and Amazon Augmented AI 5 min 🔒
4.6 Documentation: Model Cards and AI Service Cards 5 min 🔒
4.7 Veracity, safety, and the legal exposure of generated content 5 min 🔒

Module 5: Security, Compliance, and Governance for AI Solutions 14% of exam

Securing AI systems and governing the data behind them. The module starts with where the shared responsibility model draws its line for managed AI services, then works through the controls in the order an architect applies them: IAM least privilege for models, agents, and the people who build them; encryption and key management with AWS KMS; and private access through VPC endpoints and AWS PrivateLink so model traffic never crosses the public internet.From there it covers data protection — Amazon Macie for sensitive data discovery, classification, minimization, and retention — and auditability through AWS CloudTrail, Bedrock model invocation logging, and CloudWatch. The module closes on governance: data lineage and cataloging, vendor due diligence, AWS Artifact, AWS Config, AWS Audit Manager, and the external frameworks the exam names, the NIST AI Risk Management Framework and ISO/IEC 42001.

5.1 The shared responsibility model applied to AI 4 min 🔒
5.2 Identity and access management for AI workloads 5 min 🔒
5.3 Encryption and key management for AI data 5 min 🔒
5.4 Private access and network isolation 5 min 🔒
5.5 Protecting sensitive data in AI workloads 6 min 🔒
5.6 Auditability: CloudTrail, model invocation logging, and monitoring 5 min 🔒
5.7 Governance, compliance, and AI risk frameworks 5 min 🔒

Frequently asked questions about the AIF-C01 training course

Is the AIF-C01 training course free?

Yes. The course costs nothing to read and opens once you are signed in to a free Certifym account, no payment and no card.

How is the course structured?

One module per official AIF-C01 exam domain, in the order AWS publishes them, with each module weighted to the domain’s published percentage. Within each module the material is broken into short lessons, followed by key terms and further reading.

Does this replace AWS’s official training?

No. AWS publishes the official AIF-C01 exam guide and free digital courses on AWS Skill Builder, and those are the authoritative sources. This course is an independent study companion, written to be read quickly and to slot alongside practice questions, not a substitute for the official exam guide.

Do I need AI experience before starting?

The exam assumes up to six months of exposure to AI and ML technologies on AWS and familiarity with core services such as Amazon S3, AWS Lambda, and Amazon SageMaker. It does not assume you have trained a model. If AWS itself is new to you, read the CLF-C02 cloud practitioner course first, since AIF-C01 takes the underlying cloud vocabulary for granted.

What should I do after finishing the course?

Move to domain-weighted practice questions and work until you are clearing every domain consistently rather than carrying a weak module on the strength of the 28% one. The real exam is 65 questions in 90 minutes, so practice at that pace. After the exam, the natural next step is the associate-level MLA-C01 machine learning engineer course.

Is the course current with the latest AIF-C01 exam guide?

The course is written against the current exam guide, including the 2026 refresh that added agentic AI concepts: Amazon Bedrock Agents and the AgentCore runtime, Bedrock Data Automation, Prompt Management, and business-alignment metrics such as task completion rate and cost per interaction. AWS can revise the guide at any time; check the published exam guide before you book.

Trademark notice & independence. Certifym.net is operated by Certifym Exam Services, LLC and is not affiliated with, endorsed by, or sponsored by Amazon Web Services. AWS, Amazon Web Services, AWS Certified AI Practitioner, Amazon Bedrock, Amazon SageMaker, and related marks are trademarks of Amazon.com, Inc. or its affiliates. Use of these marks is solely to identify the certification for which these study materials are intended. The AIF-C01 exam guide and its objectives are the property of Amazon.com, Inc. or its affiliates; candidates should review the official, current exam guide published by AWS.

All course content, questions, answers, and explanations on Certifym are original content created for study purposes. They are not actual AWS training materials or examination questions and are not represented as such. Studying with these materials does not guarantee a passing result on any live certification exam. Exam requirements, format, domain weights, pricing, and recertification policies are set by AWS and may change; always verify current details in the published AWS exam guide before scheduling your exam.