MLA-C02 AWS Machine Learning Engineer: 5 Practice Exams
325 questions on SageMaker AI, Bedrock, RAG, MLOps and ML security, with an explanation on every answer option
Can you actually pass this thing? That is the only question a practice test exists to answer, and you should be able to answer it before you pay for a seat at Pearson VUE.
This is 325 questions for AWS Certified Machine Learning Engineer – Associate, written against the MLA-C02 exam guide published in September 2026, in a domain mix close to AWS's published weighting: Data Preparation for ML and AI 28%, ML Model and Foundation Model Development 24%, Deployment and Orchestration of ML and AI Workflows 24%, and Operating, Monitoring, and Securing ML and AI Solutions 24%.
MLA-C02 is not a re-skin of MLA-C01. The domain structure survived, but the content inside it moved: Amazon Bedrock, foundation model customization, retrieval augmented generation, agentic workflows and responsible AI now run through all four domains. MLA-C02 also dropped the ordering, matching and case study formats, leaving multiple choice and multiple response only. Practice sets written for MLA-C01 drill a blueprint and a question format the exam no longer uses.
The five tests are a deliberate ramp rather than the same test five times. Test 1 covers services and vocabulary one concept at a time. Test 2 moves to short scenarios that turn on telling neighbouring services apart. Test 3 is the trade-off test: two options are both defensible and one stated requirement decides. Test 4 goes into parameters, limits, field names and the order of operations. Test 5 is a full-length simulation with the longest scenarios, several requirements per question, domains mixed together, and the tightest clock in the set.
Every answer option carries its own explanation. Choose a wrong one and you are told what that service or setting actually does, not merely that you missed. Each question closes with an overall explanation and a link to the AWS documentation page it came from. Scores break down by the four official domain names, so a failed attempt tells you where to go back to. Pass marks are 80% for tests 1 through 4 and 85% for test 5, both set above the real exam because practising at your own desk is easier than sitting the real thing.
What this is not: a video course, a lab, or a substitute for hands-on time with SageMaker AI and Bedrock. Use it alongside the free official AWS exam prep path, and use it to decide when you are ready to book.
Groundwork: Services, Concepts, and Vocabulary
65 questions · 105 minutes · pass mark 80%
Start here, and take this test cold before you study anything. It maps the whole MLA-C02 blueprint at the level of "what is this service for and when would you reach for it", so you find out which of the four domains you already have and which one you are starting from zero on. Sixty-five questions across all four exam domains close to the published 28/24/24/24 weighting: data preparation, model and foundation model development, deployment and orchestration, and operations, monitoring and security. Questions are short and test one idea each, with no trick phrasing and no parameter values. Every answer option, right or wrong, comes with its own explanation, plus an overall explanation that links to the AWS documentation page the question came from. The pass mark is 80%, higher than the real exam, because a practice test you take at your own desk is a kinder environment than a Pearson VUE testing room, and because clearing a fundamentals test comfortably is the only sensible reason to move on to the harder ones. Allow 105 minutes. That is more time than you will get per question on exam day, which is deliberate at this stage: right now you are building recall, not practising pace.
In Practice: Picking the Right Service
65 questions · 100 minutes · pass mark 80%
The step up from knowing what a service does to knowing which of two similar services a situation calls for. Every question is a short, concrete scenario, and the distractors are the services people actually confuse in the exam room: Amazon Kinesis Data Streams against Amazon Data Firehose, AWS Glue DataBrew against SageMaker AI Data Wrangler, Amazon Macie against Amazon Comprehend, Random Cut Forest against k-means, retrieval augmented generation against fine-tuning, multi-model against multi-container endpoints. Sixty-five questions with a domain mix close to the published weighting, with per-option explanations that tell you what the wrong service actually does, so the distinction sticks instead of the answer. Includes two near-identical questions placed far apart, where a single changed requirement flips the answer. Pass mark 80%, 100 minutes.
Trade-offs: When Two Answers Look Right
65 questions · 95 minutes · pass mark 80%
The exam's favourite shape of question, and the one that separates a pass from a near miss. In every question here two options are genuinely defensible, and one requirement in the scenario decides between them: a data residency rule, a cost ceiling, a latency target, whether labelled data exists, whether the workload has to sit inside a VPC, how long a commitment can run. Learning to find that deciding requirement before you look at the options is the single most transferable exam skill, and this test drills it sixty-five times. Same domain weighting, same explanation style, per-option feedback that spells out why the runner-up loses on this specific constraint. Pass mark 80%, 95 minutes.
Under the Hood: Parameters, Limits, and APIs
65 questions · 85 minutes · pass mark 80%
Field names, parameter values, service limits, and the order operations have to happen in. This is the detail layer: the buffer interval range on Amazon Data Firehose, the record size ceiling in SageMaker AI Feature Store, the schema version string a Bedrock fine-tuning dataset needs, which auto scaling metric belongs to a real-time endpoint and which to an asynchronous one, why MaxWaitTimeInSeconds has to exceed MaxRuntimeInSeconds, where checkpoints land on a managed spot training instance. Every number in this test was verified against the AWS documentation, and figures that AWS documents inconsistently or that vary by account or Region were deliberately left out. Sixty-five questions, 85 minutes, pass mark 80%. The clock is tighter than the earlier tests because these questions are shorter to read but should be quicker to decide once you know the answer.
Exam Day: Full-Length Simulation
65 questions · 90 minutes · pass mark 85%
Save this one. Take it in a single uninterrupted sitting, no notes, no pausing, once you have worked through the first four. Every question is a full business scenario with several requirements at once, drawn from more than one domain, and you have to decide which requirement is actually load-bearing before you can answer. Sixty-five questions with a domain mix close to the published weighting, 90 minutes, and a pass mark of 85%. Both of those are set harder than the real exam on purpose. At home you are comfortable, you are not being watched, and nothing is at stake, so a score that would just scrape through in a testing centre is not evidence you are ready. Clear 85% here with time left on the clock and you can book the exam with confidence.
What you will practice
- Answer 325 MLA-C02 style questions across all four exam domains, close to the published 28/24/24/24 weighting.
- Tell apart the AWS services and features the exam deliberately confuses, from Firehose and Kinesis to RAG and fine-tuning.
- Read a scenario and spot the one requirement that decides the answer when two options both look correct.
- Recall the parameter values, service limits and API field names that the implementation questions turn on.
- Work through a rising difficulty curve, from single-concept recall to a timed full-length exam simulation.
- Learn from every attempt through an explanation on each individual answer option plus a linked AWS documentation source.
Exam domains covered
Who is this for
Developers, data engineers and IT professionals moving into machine learning engineering on AWS, and ML practitioners who want the AWS-specific certification to match the work they already do. Students and career changers preparing for their first AWS specialty-adjacent certification will find the first two tests a workable entry point, since they start at the level of what each service is for. Also useful for solutions architects, DevOps and platform engineers who now have to run and secure someone else's models, and for technical product and data leads who need to understand what their ML teams are deploying.
If you were preparing for MLA-C01, read this carefully. MLA-C01 stops being offered in English on 28 September 2026 and MLA-C02 replaces it. The four domains kept their structure but the content moved substantially: generative AI implementation, foundation models and large language models, retrieval augmented generation, agentic workflows, expanded Amazon Bedrock coverage and responsible AI practices are now spread across all four domains. MLA-C02 also dropped the ordering, matching and case study question types, so it is multiple choice and multiple response only. Practice material written for MLA-C01 prepares you for a blueprint and a question format that no longer exist.
What this is: 325 questions with a written explanation on every answer option and a documentation link on every question, arranged as a five-step difficulty curve. What it is not: a video course, a lab environment, a substitute for hands-on time in the console, and not a collection of real exam questions.
Requirements
- No prior AWS certification is required, and you can take the first test cold to find your starting point.
- Some hands-on exposure to Amazon SageMaker AI or Amazon Bedrock will help, as will basic Python and familiarity with REST APIs and AWS SDKs, since the exam profile itself assumes about a year of that experience.
- These are practice exams, not a course. They are built to run alongside the free official AWS exam prep learning path and the AWS documentation, not to replace them.
Free exam guide
Not sure what to expect? Read how the exam is built, what it covers and how to prepare.