Certification guide
MLA-C01AWS Certified Machine Learning Engineer - Associate: the honest guide
Read the date before anything else. AWS says the last day to take MLA-C01 in English is 28 September 2026. Registration for the updated version, MLA-C02, opened on 1 September 2026 as an English only beta, delivery of it begins on 29 September 2026, and the current exam stays available in Japanese, Korean and Simplified Chinese until MLA-C02 is generally available. AWS also states that your certification remains active through its full validity period regardless of when you earn it, so a pass in the final week is worth the same three years as a pass a year ago.
MLA-C01 is 65 questions in 130 minutes: 50 that count toward your score and 15 unscored items AWS uses to trial future questions, which are not marked on the exam. Results come back as a scaled score of 100 to 1,000 with 720 to pass, and you sit it at a Pearson VUE test center or online proctored.
It is the engineering exam rather than the modeling one. Data Preparation for Machine Learning is 28 percent of scored content, ML Model Development 26 percent, Deployment and Orchestration of ML Workflows 22 percent, and ML Solution Monitoring, Maintenance, and Security 24 percent. AWS's target candidate has at least a year using SageMaker and at least a year in a related role such as backend developer, DevOps developer, data engineer or data scientist.
The item types matter more here than on older AWS exams. Alongside multiple choice and multiple response, the exam guide lists ordering, matching and case study questions, and ordering and matching give credit only when every element is right.
Who ML Engineer Associate is for
A good fit if
- You already build on SageMaker and want the credential that matches the work: pipelines, training jobs, endpoints, monitoring and the access control around all of it.
- You are a data engineer or backend developer who has inherited the deployment and operations half of somebody else's models.
- You hold AWS Certified AI Practitioner. AWS lists passing the latest ML Engineer Associate exam as one of the ways to recertify it, so the two sit together deliberately.
- You can sit it in English before 28 September 2026, or you are testing in Japanese, Korean or Simplified Chinese, where MLA-C01 stays available until MLA-C02 is generally available.
- Your job title has MLOps in it, or you would like it to. The domain split is exactly the operational side of that role.
Probably not, if
- You want machine learning theory. The exam guide puts designing and architecting full end to end ML solutions, setting best practices and guiding ML strategy, and working deeply in two or more ML specialties explicitly out of scope for this candidate.
- You have never used AWS. The exam assumes a year with SageMaker plus general AWS knowledge across storage, IAM, containers, infrastructure as code and CI/CD, and none of that is taught inside the exam guide.
- You are starting from nothing this month and want the English exam. With 28 September 2026 as the last English day, the realistic routes are the MLA-C02 beta or waiting for general availability.
- You need a credential that will read as current for the next three years without a second thought. MLA-C01 is the version being replaced, and the updated exam is the one that carries the generative and agentic AI material.
Is ML Engineer Associate worth it?
For somebody already doing ML engineering on AWS, yes, and the reason is the scope rather than the badge. Two thirds of the exam is deployment, orchestration, monitoring, maintenance and security, which is the part of the work that is hardest to demonstrate in an interview and easiest to describe once an exam has forced you to name the services.
The timing changes the answer more than the content does. If you are ready now, sitting MLA-C01 before 28 September 2026 earns a certification valid for three years. If you are not, waiting is not a loss: the beta is cheaper, and AWS says the domain structure is unchanged in MLA-C02, with the additions being generative AI implementation, agentic AI, foundation models and responsible AI practices.
It is worth little as a first AWS certification. The target candidate profile assumes a year of SageMaker and a year in a related engineering role, and no amount of video course will substitute for having deployed and then debugged an endpoint.
What the exam actually asks you to do
Multiple choice and multiple response, plus the newer AWS item types: ordering, matching, and case studies with several questions on one scenario.
Item formats
- Multiple choice
- Multiple response
- Drag and drop
The highlighted formats are the ones you cannot answer from memory alone. AWS, ML Engineer Associate exam guide ↗
Domain breakdown and official weightings
From the official AWS exam guides. Data Preparation for Machine Learning is the heaviest domain at 28 percent, followed by ML Model Development at 26 percent.
- Data Preparation for Machine Learning28%
- ML Model Development26%
- Deployment and Orchestration of ML Workflows22%
- ML Solution Monitoring, Maintenance, and Security24%
Study plans by experience level
Working ML engineer on AWS
3 to 5 weeksat 6 to 8 hours
- 1Read the exam guide end to end, including the appendix that names in scope and out of scope services. On this exam the appendix is the fastest way to find what you have never touched.
- 2Work the four domains in weighting order and be honest about the second one. Data preparation at 28 percent covers formats, ingestion, feature engineering, encoding, bias detection and data integrity, and engineers who joined after the pipeline was built often know the least about it.
- 3Take the official practice question set early, not late, to learn the ordering, matching and case study formats. Losing an ordering question because you had the right four steps in the wrong sequence is an avoidable way to fail.
- 4Spend a session each on the monitoring and security domain's specifics: model and data drift detection, logging and troubleshooting, cost analysis, and the identity, encryption and network isolation options around endpoints. It is 24 percent and it is the domain most often studied last.
- 5Book the sitting before 28 September 2026 and leave 14 days of slack in case the first attempt does not go your way.
Data or backend engineer new to SageMaker
10 to 12 weeksat 8 to 10 hours
- 1Build one small model end to end in your own account before reading any exam material: ingest data to S3, transform it, train, register the model, deploy an endpoint, invoke it, then delete the endpoint. Everything the exam asks about is a variation on that path.
- 2Learn the deployment shapes and when each applies: real time endpoints, serverless endpoints, asynchronous endpoints and batch inference, plus the tradeoffs in latency, cost and throughput that decide between them.
- 3Move to orchestration: SageMaker Pipelines, CI/CD with the AWS developer tools, infrastructure as code, containers, and blue green, canary and linear deployment strategies with the rollback that goes with each.
- 4Cover model development at the level the exam actually asks for: choosing between built-in algorithms, foundation models and AWS AI services, hyperparameter tuning, and the evaluation metrics, since interpreting a confusion matrix or an AUC score is examinable and deriving one is not.
- 5Finish on monitoring and security, then take the official pretest under time and use its section feedback to decide what the last fortnight is for.
AI Practitioner holder moving up
6 to 8 weeksat 6 to 8 hours
- 1Accept that the vocabulary carries over and the depth does not. AI Practitioner asks what a service is for; this exam asks which instance type, which endpoint shape and which scaling policy, and expects you to have configured them.
- 2Start with the hands-on gap. Provision an endpoint, script it with infrastructure as code, put automatic scaling on it and watch the metrics move, because the exam's deployment and orchestration domain is written for people who have done that.
- 3Work data preparation next. It is the largest domain and the one furthest from the AI Practitioner syllabus: formats, ingestion, feature engineering, labeling and data integrity rather than concepts.
- 4Read the exam guide's task statements as a checklist and mark each one as done, read about or never touched. Convert the third category into lab time rather than reading time.
- 5Use the official practice question set for the newer item formats, then book before 28 September 2026 if you want the English exam.
Common mistakes
- Studying the model and skipping the pipeline
- Deployment and orchestration plus monitoring, maintenance and security are 46 percent of scored content between them, and model development is 26 percent. Candidates who come from a data science background revise algorithms and lose the exam on endpoints, scaling policies, CI/CD and drift detection.
- Ignoring the newer item types
- Ordering questions give credit only when three to five steps are in the correct sequence, and matching questions only when every pair is right. A case study puts several questions on one scenario. Meeting any of the three for the first time on exam day costs marks that revision cannot get back.
- Pacing against 50 questions instead of 65
- The exam includes 15 unscored questions that are not identified, so you answer 65 items in 130 minutes. That is two minutes each, and the case studies take longer to read than a standard question, so the pace has to come from somewhere else.
- Leaving lab resources running
- The services this exam is about are metered. An endpoint or a training job left provisioned overnight can cost more than the exam fee. Set a billing alarm before the first notebook and delete each resource the same day you finish with it.
- Booking the English exam without checking the date
- 28 September 2026 is the last day to take MLA-C01 in English, and a failed attempt carries a 14 calendar day wait before a retake. A first sitting booked in the last fortnight leaves no room for a second, and after the deadline the English route is the MLA-C02 beta rather than this exam.
- Treating the security domain as somebody else's job
- ML Solution Monitoring, Maintenance, and Security is 24 percent, and it asks about IAM roles and policies for ML systems, network isolation for endpoints, encryption and compliance features. On a real team that work often sits with a platform group, which is exactly why candidates arrive without it.
What comes after passing
The certification is valid for three years, and AWS lists two ways to renew this one: pass the latest version of the exam, or pass the AWS Certified Generative AI Developer Professional exam. Time is added from the date you complete the recertification rather than from the old expiry, so there is no reward for renewing very early.
Look in your AWS Certification Account before you book anything else. AWS names the 50 percent discount voucher there as the way to pay for a recertification exam, and it is issued for passing rather than for failing.
Follow where the role is going rather than where the exam was. AWS says MLA-C02 keeps the same four domains and adds generative AI implementation, agentic AI orchestration, foundation model selection and fine tuning, and responsible AI practices, which is a fair description of what ML engineering on AWS is turning into.
Where people go next
Costs across the full renewal cycle are on the ML Engineer Associate cost page.
Frequently asked questions
When does MLA-C01 retire?
AWS says the last day to take the current exam in English is 28 September 2026. It remains available in Korean, Japanese and Simplified Chinese until MLA-C02 reaches general availability, and a certification you already hold stays active through its full three year validity period.
How much does the MLA-C01 exam cost?
$150, which is the AWS price for every Associate level exam. The updated MLA-C02 version is priced as a beta at $75, is 85 questions in 170 minutes, opened for registration on 1 September 2026 in English only, and can be taken only once.
What score do I need to pass MLA-C01?
720 on a scale of 100 to 1,000. The exam includes 50 scored questions and 15 unscored ones that do not affect your score, and AWS uses a compensatory model, so you need to pass the exam overall rather than each section.
What is the AWS retake policy?
If you fail, you must wait 14 calendar days before you are eligible to retake the exam. There is no limit on the number of attempts, but you pay the full registration fee each time. Once you have passed, you cannot retake the same exam for two years unless it is updated with a new exam guide and exam series code.
Should I take MLA-C01 or wait for MLA-C02?
If you are ready before 28 September 2026, MLA-C01 earns the same certification for the same three years. If you are not, AWS says the updated exam keeps the same four domains and adds generative AI, agentic AI, foundation model and responsible AI content, so waiting costs you nothing in scope and gains you the current syllabus.
Keep reading
- ML Engineer Associate practice questions
Free sample questions with the full explanation on every answer.
- Free ML Engineer Associate practice test
10 original questions, playable now. No account, no card.
- ML Engineer Associate exam format
How many questions, how long, what the items look like, and how long it stays valid.
- ML Engineer Associate passing score
The exact cut score, what kind of number it is, and the retake terms.
- How hard is ML Engineer Associate?
An honest difficulty read from the format, the clock and the weights.
- What ML Engineer Associate costs
The voucher price, the retake, and what renewal costs across the cycle.
Practice ML Engineer Associate for free while you decide
Original questions written from the published objectives, with the concept, the reasoning, and a note on every wrong option. No account needed to start.
Start free ML Engineer Associate questions