Book the AWS Certified Machine Learning Engineer exam in English today and you’ll sit the MLA-C02 beta, a version we found no major-publisher study guide for. English delivery of MLA-C01 ended on September 28, 2026, and AWS retires it in every language on January 14, 2027, when MLA-C02 reaches general availability. The AI Practitioner certification kept its AIF-C01 code. AWS did revise its exam guide in April 2026 to expand its agentic AI coverage, though, so all the AWS AI Practitioner books we found predate part of what is now in scope.
We built this list around that gap. For AIF-C01 you still want a study guide built around its five domains, plus a book on agents for the April additions. For the ML Engineer certification, the honest stack is an MLA-C01 guide for the four domains AWS kept and a 2026 book on Bedrock, RAG and agents for the parts it added. We checked the certification details against AWS’s own certification pages and exam guides, and the editions, prices and stock against Amazon’s catalog, on October 19, 2026.
The seven picks at a glance
Each pick earns its place for one job. Most readers need two books: one from the top three rows and one from the rest.
| Award | Book | Exam fit | Why it wins | Price |
|---|---|---|---|---|
| Best for AIF-C01 | AWS Certified AI Practitioner Study Guide (Sybex) | AIF-C01 | Domain-shaped chapters plus an online bank with 2 practice exams | about $49 |
| Best shorter AIF-C01 read | AWS Certified AI Practitioner (AIF-C01) Study Guide (O’Reilly) | AIF-C01 | A tighter read with walkthroughs of Bedrock and SageMaker | about $53 |
| Best for the ML Engineer core | AWS Certified Machine Learning Engineer Study Guide (Sybex) | MLA-C01, core of MLA-C02 | The only major-publisher MLA guide in print, with a test bank | about $45 |
| Best bridge to MLA-C02 | Machine Learning Engineering on AWS, 2nd Edition | MLA-C02 additions | Bedrock, AgentCore and LLMOps pipelines, published May 2026 | about $50 |
| Best for agents on AWS | AI Agents on AWS | Both | Strands, AgentCore, MCP and Guardrails in one cheap book | about $31 |
| Best vendor-neutral foundation | AI Engineering | Both | Evaluation, RAG, agents and fine-tuning explained properly | about $52 |
| Best for model internals | Generative AI on AWS | Background | Fine-tuning, PEFT, RLHF and deployment on AWS | about $43 |
What changed in both exams this year
Two separate changes hit these certifications in 2026, and they age books in different ways. AIF-C01 changed in place: AWS published version 1.1 of its exam guide on April 30, 2026, and says guide updates reach the live test about a month after they are published. MLA-C01 is being replaced outright by a new version with a new code, which is why a book’s cover code matters more for the ML Engineer certification than for AI Practitioner.
AWS’s April revision expanded the agent and application topics, and named services that no 2025 study guide was written against:
- Agentic AI concepts: multi-agent patterns, the Model Context Protocol (MCP), agent memory, tool use and workflow orchestration (MCP was already named in the earlier guide; version 1.1 gives these their own objective).
- Context engineering and token pricing: how context shapes a foundation model’s output and how token-based billing works.
- New services in scope: Amazon Bedrock AgentCore, Strands Agents, Kiro, Amazon Q, SageMaker JumpStart, AWS Transform and Aurora. MemoryDB dropped out.
- Governance and application additions: Bedrock Prompt Management for prompt versioning, AgentCore Identity and Policy, Bedrock Guardrails, and hallucination detection and grounding.
The rest of AIF-C01 is unchanged: 65 questions (50 scored) in 90 minutes, a 700 passing score, $100, and the same five domains, led by Applications of Foundation Models at 28%. Kiro, AWS’s AI coding IDE, is now in scope, and our comparison of Cursor, Windsurf and Kiro shows what that tool does in practice. The AWS AI Practitioner certification page has the current languages; AWS retired the Italian and German versions after October 15, 2026.
For the ML Engineer certification, the change is bigger but less disruptive than the new code suggests. AWS kept the four domains and reweighted them from 28/26/22/24 to 28/24/24/24, so data preparation still leads and deployment gains two points. What it added is a generative AI layer across all four, and it removed a few C01 skills:
- Added: vector databases, document chunking for RAG, Bedrock foundation model selection and knowledge bases, agent deployment and state, LLM-as-a-judge evaluation, metrics such as BLEU and ROUGE, Bedrock Guardrails and token cost patterns.
- Removed: SageMaker Neo edge optimization, bring-your-own-container on SageMaker, loading training data from EFS or FSx, model size reduction, and the C01 skill on fine-tuning pre-trained models with custom datasets. Fine-tuning itself stays in Domain 2, reframed around foundation models, and preparing fine-tuning data is now a Domain 1 skill.
- Format: the beta (code ME1-C02) runs 85 questions in 170 minutes for $75 in English only; the standard version scores 50 questions and keeps the 720 passing score.
AWS set the dates in its September 2026 certification announcement: beta delivery from September 29, 2026, and general availability in all languages on January 14, 2027 (the certification page itself still lists GA as TBD), the same day MLA-C01 retires in Japanese, Korean and Simplified Chinese. The older Machine Learning Specialty (MLS-C01) is already gone, with a last test date of March 31, 2026, so skip any book with that code on the cover.
How we picked
We only considered books from publishers with an editorial process (Wiley’s Sybex line, O’Reilly and Packt), in print and buyable new on October 19, 2026. Every page count, publication date and price comes from Amazon’s catalog API that day. Every chapter claim comes from the publisher’s page or cover, the book’s official GitHub repository, or (for the O’Reilly AIF-C01 guide) O’Reilly’s description as syndicated to Apple Books. We read those tables of contents against both AWS exam guides. We did not sit either certification to grade the books, so the fit you see below is coverage, not a pass rate.
We left three groups out on purpose:
- Packt’s MLA-C01 study guide: a preorder that ships December 9, 2026, five weeks before the code it targets retires.
- Independently published titles: indie AIF-C01 guides have been on sale since 2025, and several MLA-C02 titles appeared within weeks of AWS’s July 2026 announcement, one sold on the promise of “real exam-style questions”.
- Machine Learning Specialty books: the MLS-C01 certification retired in March 2026.
Compare the seven books
Publication date is the column to read first, because it tells you which revision of each certification a book could have seen. Any study guide published before April 30, 2026 predates the AIF-C01 April additions, and nothing here was written against the MLA-C02 guide:
| Book | Published | Pages | Practice questions | Agents coverage |
|---|---|---|---|---|
| Sybex AI Practitioner Study Guide | October 2025 | 320 | Online bank: 2 practice exams, flashcards | RAG and agents chapter |
| O’Reilly AIF-C01 Study Guide | September 2025 | 275 | In the book | Not confirmed |
| Sybex ML Engineer Study Guide | June 2025 | 448 | Online bank: assessment test, practice exam, flashcards | None |
| Machine Learning Engineering on AWS, 2nd Ed. | May 2026 | 548 | None | AgentCore, Strands, knowledge bases |
| AI Agents on AWS | April 2026 | 290 | None | The whole book |
| AI Engineering | January 2025 | 532 | None | RAG and agents chapter, no AWS |
| Generative AI on AWS | December 2023 | 309 | None | RAG and agents chapter, pre-AgentCore |
Put side by side, the books split cleanly: the two practice banks sit in books written before the April agent additions, and the AgentCore and Strands material sits in books with no practice questions. That is why pairing a Sybex guide with one of the 2026 books covers more than buying two study guides for the same code.
1. AWS Certified AI Practitioner Study Guide (Sybex)
This is the AIF-C01 book to buy if you only buy one, because its structure follows the certification’s domains and it comes with the practice bank the others lack. All three authors are AWS specialists: Vikram Elango is a generative AI specialist solutions architect, Shreyas Subramanian a principal data scientist, and Vivek Gangasani leads inference solutions architecture.
Its ten chapters run from AI and generative AI basics through AWS’s AI services, model selection and prompt engineering, a chapter on RAG and agents, model customization, MLOps, responsible AI, and security and governance. Those last two map onto the two domains weighted at 14% each. The book includes a year of access to Sybex’s online learning environment, which holds two custom practice exams, chapter review questions, more than 100 flashcards and a searchable glossary.
October 2025 is its weak point. The book shipped before the April 2026 additions (AgentCore, Strands Agents, Kiro, MCP and context engineering), and Wiley’s table of contents has no section on any of them, though chapter 6 covers Bedrock Agents and multi-agent collaboration and chapter 9 covers Bedrock Guardrails and LLM-as-a-judge. Cover the rest from AWS’s own guide or pick 5. On October 19 the buy box belonged to a third-party seller rather than Amazon, so check who is shipping before you order.
Best for: anyone sitting AIF-C01 who wants a domain-by-domain guide and timed practice. Skip it if AgentCore, Strands and Kiro need to be in print, because this edition predates them. It sells for about $49 against a $63 list price; check the current price.
2. AWS Certified AI Practitioner (AIF-C01) Study Guide (O’Reilly)
Tom Taulli’s guide is the shorter route to the same certification, at 275 pages against Sybex’s 320. It suits readers who learn better from plain explanation than from a test bank. O’Reilly’s description puts machine learning, natural language processing, computer vision and generative AI on the syllabus, with hands-on time in SageMaker, Bedrock, Rekognition and Comprehend.
Practice questions and exam-focused strategies are in the book itself, but O’Reilly does not advertise a separate online test bank, and that’s the real difference from pick 1. We could not confirm how much agent material it has, because O’Reilly’s catalog pages block automated access and the description does not mention agents. The print edition came out in September 2025, so treat it like the Sybex guide: solid on the five domains, not current on the April revision.
Best for: readers who want a readable AIF-C01 guide sold and shipped by Amazon. Pass on it if the Sybex guide is already on your desk, or full mock exams matter, since this one does not bundle them. Expect about $53; check the current price.
3. AWS Certified Machine Learning Engineer Study Guide (Sybex)
Dario Cabianca’s guide is the only major-publisher ML Engineer study guide in print, and it still earns a place even though its cover says MLA-C01. AWS kept all four domains for MLA-C02. This book’s eight chapters track them closely: data ingestion and storage, transformation and feature engineering, model selection, training and evaluation, deployment and orchestration, monitoring and cost optimization, and security.
The online test bank adds an assessment test, chapter review questions, a practice exam and flashcards, and Appendix B covers the math essentials. What it does not have is any RAG, Bedrock or agents chapter, which is where most of the MLA-C02 additions live. Pair it with pick 4 or pick 5, and use our MLOps book picks if the deployment and monitoring chapters leave you wanting more depth.
Best for: ML Engineer candidates who want the four core domains drilled with practice questions. Skip it if generative AI is all you are after. We found no Sybex MLA-C02 edition listed on October 19, so waiting for one is a gamble. It ran about $45 from a third-party seller on October 19; check the current price.
4. Machine Learning Engineering on AWS, 2nd Edition
Joshua Arvin Lat’s second edition, published in May 2026, is the closest thing in print to a study companion for what MLA-C02 added. It is not an exam guide. Its opening chapters build agents with Strands Agents, Bedrock foundation models, Bedrock Knowledge Bases, S3 Vectors and AgentCore Runtime before the book moves into SageMaker AI.
From chapter 3 onward it’s the traditional ML engineering half: a transactional data lake, data management and processing on AWS, SageMaker AI training and tuning, deployment options and strategies, and LLMOps pipelines with SageMaker Pipelines. That spread lines up with the RAG, knowledge base and agent deployment tasks AWS added, and with the pipeline work that was already in MLA-C01. You will want an AWS account with Bedrock access before chapter 1, and our AWS CLI setup guide covers the local tooling.
Best for: engineers who learn by building and want Bedrock and AgentCore in their hands before MLA-C02. Not for: readers who need domain mapping and practice questions. Avoid the 2022 first edition, which sells for almost the same price and predates the generative AI chapters. It sells for about $50; check the current price.
5. AI Agents on AWS
Bunny Kaushik and Mona M wrote the cheapest way to cover the agent topics AWS added to both certifications this year, at about $31 for 290 pages. It was published in April 2026, and its code runs on Bedrock, AgentCore and the Strands Agents SDK.
Chapter by chapter, it follows the order you’d build an agent in, and most chapters line up with a topic AWS added:
- First agents and tools: Strands and LangGraph basics, then agents that call tools.
- Memory: Mem0 and AgentCore Memory, which covers the agent memory and state topics.
- Multi-agent patterns: supervisor and worker, agent as a tool, graph routing and swarm.
- Protocols: MCP and agent-to-agent (A2A) communication.
- Production: deployment to AgentCore Runtime, ECS and Lambda, then evaluation with LLM-as-a-judge, OpenTelemetry tracing, Bedrock Guardrails and AgentCore Policy.
For an AIF-C01 candidate this is the fastest way to make the April additions concrete; for MLA-C02 it covers agent deployment, state and monitoring with working code. Our wider AI agents book guide has framework-neutral options if you want theory first.
Best for: anyone who wants hands-on agent code on AWS for about $31. Skip it if Python is new to you, or AIF-C01 is your only goal and the concepts in AWS’s own exam guide are enough. Check the current price.
6. AI Engineering
Chip Huyen’s book has no AWS content at all. It is still on this list because it explains the ideas both exam guides keep returning to better than any AWS-specific title. It held an Amazon Best Sellers Rank of 2,274 on October 19, far ahead of every other book here, all of which ranked outside the top 100,000.
Two of its ten chapters cover evaluation, the area where MLA-C02 now asks about LLM-as-a-judge and metrics such as ROUGE. The rest take prompt engineering, RAG and agents, fine-tuning, dataset engineering and inference optimization. Huyen says plainly in the book’s repository that it is not a tutorial and has few code snippets. That makes it the reference you return to after the study guide, and our LLM engineering book picks list where to go next.
Best for: the reasoning behind foundation model applications, whichever cloud you run them on. Look elsewhere if a test date is close and AWS service names and console steps are what you need. It sells for about $52; check the current price.
7. Generative AI on AWS
Chris Fregly, Antje Barth and Shelbee Eigenbrode wrote this one in December 2023, and it is the deepest book here on how foundation models are trained and tuned on AWS. Its twelve chapters cover the project lifecycle, prompt engineering, quantization and distributed computing, fine-tuning and evaluation, PEFT, RLHF, deployment, RAG and agents, multimodal models, Stable Diffusion and a closing chapter on Bedrock.
Bedrock is where its age shows: the book was written before AgentCore and Strands Agents existed, so read its single Bedrock chapter as history and take the current service list from picks 4 and 5. Its value for MLA-C02 is the model development domain, where fine-tuning and evaluation still need the depth this book gives. At about $43 against a $79.99 list price, it is also the cheapest of the three O’Reilly titles here.
Best for: ML Engineer candidates who want to understand fine-tuning and model optimization, not just name the service. Skip it if current Bedrock features are the goal, since one 2023 chapter cannot cover them. Check the current price.
What to check before buying AWS AI Practitioner books
The AWS AI shelf on Amazon mixes current guides with retired codes and self-published titles, and a few checks sort them before you pay:
- The code on the cover: AIF-C01 is current; MLA-C01 retires on January 14, 2027 and MLS-C01 already has. A C01 ML Engineer guide is still useful for the core domains, but only paired with a 2026 book.
- The publication date: any study guide published before April 30, 2026 predates the AIF-C01 April additions, and nothing from a major publisher was written against the MLA-C02 guide yet.
- The practice bank: only the two Sybex guides advertise online mock exams; the build books have none.
- The seller: both Sybex guides were sold by third parties on October 19, so check the seller and the return terms before you buy.
- The publisher: self-published MLA-C02 titles appeared within three months of AWS’s July announcement. We could not verify who wrote them or how, so we left them out.
The same filters work for AWS’s other certifications, and our Solutions Architect Associate book guide applies them to SAA-C03, where retired-code books are just as common.
The study budget for each path
On either track the books cost between half and about a quarter more than the certification fee, so the cheapest plan is the one you pass first time. These totals use the October 19 prices above and the fees on AWS’s certification pages:
| Path | Exam fee | Books | Total |
|---|---|---|---|
| AIF-C01, guide only | $100 | Sybex AI Practitioner (about $49) | about $149 |
| AIF-C01 with agents | $100 | Sybex AI Practitioner + AI Agents on AWS (about $80) | about $180 |
| ML Engineer, MLA-C02 beta (while the beta runs) | $75 | Sybex ML Engineer + ML Engineering on AWS (about $95) | about $170 |
| ML Engineer, MLA-C02 at general availability | Not yet listed | Same two books (about $95) | Fee plus about $95 |
At $75 the beta is the cheaper sitting, but it is English only and runs 85 questions in 170 minutes, which is a long afternoon for a first attempt. AWS has not printed a general-availability fee for MLA-C02 on its certification page yet; MLA-C01 cost $150, so budget for that until AWS says otherwise. You can sit the beta only once; if you fail, the retake waits for the general-availability version, and AWS says betas generally run one to three months, so the window may close before January 14; book soon if you want the $75 price.
If you plan to hold both, take AIF-C01 first. AWS says earning the ML Engineer Associate automatically recertifies an existing AI Practitioner certification, so passing MLA-C02 later renews AIF-C01 without a second sitting. If you are paying for the lab time yourself, our breakdown of real AWS costs shows what a Bedrock or SageMaker practice account runs to before the bill surprises you.
