AWS Certified Machine Learning Engineer - Associate: What to Know Before You Register

What This MLA-C01 Exam Actually Certifies

The AWS Certified Machine Learning Engineer - Associate exam checks whether you can take a machine learning workload from raw data to a running, monitored production system on AWS. It is not testing whether you can explain gradient descent from a whiteboard. It is testing whether you know how to get data ready for SageMaker, get a model trained and deployed, wire up a pipeline that redeploys it automatically, and keep the whole thing observable and secure once it's live. AWS is explicit that designing a full ML architecture from scratch, or going deep into a specialized area like computer vision or NLP, sits outside the scope of this exam. This is a builder and operator credential, not an architect or researcher one.

Who This MLA-C01 Exam Is For (and Who It Isn't)

AWS aims this at people already doing the job in some form: backend developers who've been pulled into ML pipeline work, DevOps engineers standing up SageMaker infrastructure, data engineers feeding models, and data scientists who need to operationalize what they build. The target candidate has roughly a year of hands on time with SageMaker and related AWS services.

If you've never touched SageMaker, an IAM policy, or a CI/CD pipeline, this exam will feel like a steep climb regardless of how strong your general ML theory is. Conversely, if you have deep ML research experience but haven't deployed anything on AWS, the AWS-specific mechanics (endpoint types, pipeline orchestration tools, monitoring services) will be the gap, not the ML concepts themselves.

MLA-C01 Exam Format and What Taking It Feels Like

You'll sit for 130 minutes with 65 questions, of which 50 count toward your score and 15 are unscored items AWS is quietly testing for future exams. You won't know which is which, so every question deserves full attention.

Most questions are multiple choice (one correct answer, three distractors) or multiple response (select all that apply from five or more options, no partial credit). You'll also run into ordering questions, where you sequence 3 to 5 steps correctly, and matching questions, where you pair items from two lists. These formats reward candidates who've actually built pipelines end to end, because you need to know not just the right service but the right order of operations.

Scenario length and specificity tend to be the time pressure. A typical question describes a business situation in a paragraph or two, then asks which combination of services or configuration choices solves it, often with two plausible sounding answers that differ on a subtle AWS behavior. Reading carefully costs you more time than the ML knowledge itself.

Domains and Their Weightings

AWS publishes four content domains for MLA-C01, and this weighting is taken directly from the current exam guide:

Data preparation carries the single heaviest weighting, which surprises candidates who expect model building to dominate. In practice this means you should be comfortable with data ingestion and transformation services well before you get deep into algorithm selection.

Cost Breakdown

MLA-C01 costs 150 USD to sit, with pricing in other currencies varying by country. If you don't pass, AWS requires a wait before you can retake, and you pay the full fee again each attempt; there's no reduced retake price. AWS occasionally runs regional promotions or half-price voucher offers through its own channels, so it's worth checking your AWS Training and Certification account before assuming you'll pay full price.

The successor exam, MLA-C02, is priced at 75 USD during its beta period. Beta pricing is typically temporary and tends to rise once an exam moves to general availability, so treat that figure as a limited-time entry point rather than the long-term cost.

Prerequisites and Eligibility

There's no exam or credential you're required to hold before registering. AWS recommends a year of hands-on SageMaker experience and a year in an adjacent technical role, but nothing stops you from registering without either. What actually gates your success is whether you've built things: an ingestion pipeline, a trained and deployed model, a monitoring setup. Reading alone rarely closes that gap.

How the MLA-C01 Exam Is Scored

Results come back as pass or fail, with a scaled score between 100 and 1,000 and a passing threshold of 720. AWS uses a compensatory model, meaning your overall score is what matters, not clearing a minimum bar in each domain individually. You could be weak on deployment and orchestration and still pass if data preparation and model development pull your average up. Your score report will show a breakdown by section, but AWS itself cautions against reading too much into that section-level detail since it's meant as general feedback, not a precise diagnostic.

Renewal and Recertification

The certification stays valid for 3 years from the date you earn it. Recertifying means passing whatever the current version of the exam is at that point, which practically speaking means passing MLA-C02 or whatever has succeeded it by your renewal date, not necessarily MLA-C01 again.

The Transition You Need to Know About Right Now

This is the part of the page that changes the calculus for anyone reading it in the next few weeks. AWS is retiring MLA-C01 in English on September 28, 2026. If your target sitting date is on or before that, you're taking the current exam and everything above applies as written. If you're registering after that date, English candidates will be sitting MLA-C02 instead, which is currently in a beta window that opened September 1, 2026 and began beta delivery on September 29, 2026.

MLA-C01 remains available in Japanese, Korean, and Simplified Chinese until MLA-C02 reaches general availability in those languages, so non-English candidates have more runway than English-language candidates do.

What's actually changing in MLA-C02: the four content domains stay the same, but AWS has broadened the task statements to cover generative AI implementation using Amazon Bedrock and retrieval-augmented generation patterns, agentic AI orchestration, and foundation model selection and fine-tuning, on top of the existing traditional ML content. The exam also grows to 85 questions across 170 minutes, up from 65 questions and 130 minutes. If you were already deep into an MLA-C01 study plan built around traditional SageMaker workflows, budget extra time to cover the generative AI and agentic AI additions before sitting MLA-C02.

One practical wrinkle: beta exam results typically take about 5 business days to come back, rather than the immediate pass/fail you'd normally get, because AWS is still calibrating which questions will count toward scoring going forward.

How This MLA-C01 Exam Compares to Its Closest Alternative

The natural point of comparison used to be AWS Certified Machine Learning - Specialty (MLS-C01), but that exam was fully retired on March 31, 2026, so it's no longer an option to compare against as a live alternative. AWS has been explicit that it's moving away from a model-first, algorithm-heavy specialty exam toward role-based, engineering-focused credentials like this one.

If you want something broader and less technical as a starting point, AWS Certified AI Practitioner is the foundational-level credential covering AI and ML concepts without the hands-on engineering depth this exam demands. If you're aiming higher, AWS has introduced a Generative AI Developer - Professional credential aimed at production-focused generative AI work, which sits above this Associate-level exam in both scope and expected experience.

Outside the AWS ecosystem, the closest equivalents from other cloud providers are Microsoft's Azure AI Engineer Associate and Google's Professional Machine Learning Engineer, both of which cover similar ground (deploying and operationalizing ML on a specific cloud) but obviously test against that provider's own services instead of SageMaker and Bedrock.

Preparation Timeline

For someone already working hands-on with SageMaker regularly, 4 to 6 weeks of focused review covering the domains you use least in your day job is a reasonable target. If you're coming from adjacent experience, like general cloud engineering or data science without much AWS-specific deployment work, 8 to 12 weeks is more, with the bulk of that time going toward actually building something end to end in a SageMaker environment rather than just reading about it.

Given the current transition, timeline planning also means picking a lane: if you can reach exam readiness before September 28, 2026, sitting MLA-C01 now means a shorter, narrower exam. If that date isn't achievable, plan your prep around MLA-C02's expanded scope from the start rather than cramming generative AI and agentic AI content in later.