Everything you need to know about AWS Machine Learning Specialty Exam

Everything you need to know about AWS Machine Learning Specialty Exam


15 min read

If you are going to take the AWS Certified Machine Learning Specialty exam, you might have a lot of questions regarding the things you need to know before taking the $300 exam.

I passed the exam in July 2022, and here I will be writing a comprehensive guide on what the exam entails along with some additional info, like the benefits of having the certificate, how to save on certificate expenses, and advice for the exam session. However, I am not going to provide the answer to the questions that I saw during the exam, as that would compromise the integrity of taking the exam. If I were to go back in time, I would follow this guide, as it includes everything that worked/didn’t work for me.

But if you’re interested in one particular part, you can always use the anchor links in the table of contents.

Why go for the certificate? (or is it worth it?)

You will see a lot of reasons listed here and there (educational websites, data blogs, etc.), on why you should take this exam. These might include higher pay, more credibility, popularity, etc. Although these are not false, to be realistic, passing the exam is not going to make you a Machine Learning hero overnight.

In my opinion, two main logical reasons for taking the exam are:

  • You’re looking for a job, and you want to showcase your ML skills on AWS.

  • Your current employer needs you to have the certificate in order to collaborate with other companies.

AWS also offers a few benefits in case you pass the exam:

  • You get a 50% discount on your next certification exam (quite useful)

  • You get access to buying shirts and other gear that says AWS-certified! (interesting for some, but no for others)

  • You get to join the Global AWS Certified Community.

  • You get to join the Subject Matter Expert (SME) Program, which allows you to review certification exams, which in turn, might lead to recognition and some items from the store!

My intention for this paragraph was to give you a clear and realistic view of what the certificate can do for you. Now let’s get down to what it takes to pass the certification exam.

What does the exam entail?

The exam is a combination of 65 multiple choice/multiple answer questions. In case the question has multiple answers, you will receive instructions on how many you need to choose.

Overall, four general domains will be covered in the exam. Here is a list with a breakdown of their appearance in the exam:

  • Data Engineering (20%: ~13 questions)

  • Exploratory Data Analysis (24%: ~16 questions)

  • Modeling (36%: ~23 questions)

  • ML Implementation & Operations (20%: ~13 questions)

It is important to know that the exam is not exclusively about ML on AWS, but it is centered around Machine Learning in general, and how it can be done using AWS.

Data Engineering

This part focuses on data preparation practices and specifically includes AWS storage services (mainly S3), data streaming services (mainly Kinesis), ETL services (Glue, Athena, EMR), and coordination services (AWS Pipeline, Step Functions, etc).

Exploratory Data Analysis

In general data science projects, this stage is when you have the preprocessed data ready and start doing analytics to get insights. The exam will test your knowledge of differentiating between different kinds of analytics, visualizations, and challenges you might face in real-life settings (outliers, missing values, etc).


This is the main part of the exam, focusing on data science and machine learning both in general and the tools to do modeling on AWS. This part comprises more than a third of the whole exam. Since this area is broad, we’re gonna break it down into smaller sub-domains.

1. General Machine Learning

This part is about the general foundation of Machine Learning: regularization, training metrics and where each of them is an optimal choice, deep learning fundamentals like activation functions, convolutions, recurrent neural nets, etc. If you want to take the ML Specialty Exam there’s a good chance that you already know these concepts, as you most probably have trained models before, but you would definitely want to brush up on them.

2. SageMaker

Well, it is the most sophisticated service that Amazon offers for AI and Machine Learning, so it would actually make sense that a good portion of the exam is dedicated to it. There are two main subsections in this part. SageMaker services and tools, and the built-in algorithms.

SageMaker Tools

Studio, canvas, feature store, Notebook instances and their configurations, Autopilot, Hyperparameter Tuning, Training jobs, GroundTruth, Model Monitor, Neo, etc. You have to know about them all, where they are useful and how you can configure them for a specific job.

SageMaker Built-in Algorithms

You can see the full list of SageMaker Built-in Algorithms here. To successfully answer these questions, not only do you need to know what each of them does and where they can be useful, but you also need to know the following:

  • What are the valid data input types

  • The types of instances you can use for training

  • The types of instances you can use for inference

  • Whether they support GPUs or not

  • Whether they support distributed training or not

3. High-level SaaS ML Services

This part is dedicated to high-level ready-for-inference Software-as-a-Service (SaaS) machine learning products. These include services like Comprehend, Rekognition, Translate, Forecast, etc, that are powered by trained machine learning models. For this part, you have to know which is helpful and when. Sometimes the question asks which option would achieve some task, “with the least effort” or “quickest way possible”. When you see these phrases, always look for SaaS, and put them as a higher priority while investigating other options (don’t jump to conclusions quickly)!

ML Implementation and Operations

This part is more inclined toward the ML Engineers and DevOps Specialists than data scientists. Having hands-on experience with AWS can be beneficial here. The questions revolve around best practices of productizing ML models on AWS infrastructure, balancing loads, making highly-available and resilient architectures, pipelining your workflow, and applying security measures.

So now let’s get to the main point that is the most probable reason you’re reading this article.

How to prepare for the exam

This exam is not about your machine learning/engineering/data science knowledge. So never assume you can pass it, solely relying on your ML knowledge or AWS experience. I would like to think about this exam as International English Language Testing System (IELTS), where even native English speakers can score poorly if they don’t know what they will be measured for.

AWS Resources

The official webpage for AWS Certified Machine Learning Specialty Exam can be a good starting point. At the bottom, you can find links to some sample questions, webinars, white papers and the AWS Skill Builder.

These resources are quite informative, but the information is scattered and overall is not the fastest way towards passing the exam. If you can take your time with studying all the material, you might not even need external resources. However, the exam at around €300 is an expensive one, and if you don’t feel confident you would be able to pass it in a short time (as was the case with me), you can spend an extra couple of bucks on some external resources.

External Resources


The most helpful resource I found on the internet was this Udemy course. It is well-structured, regularly updated and maintained. They cut to the chase and will focus on the types of questions that might show up on the exam.

The original price is a whopping €170, but fortunately and like many other Udemy courses, there is an on-going discount which enables you to buy the course for an affordable €20 price. I think it is definitely worth it.

A couple of sample questions are available with this course, but for a full sample exam, you need to pay another €20 to buy this course. I didn’t, because I found cheaper and more extensive exam sets in…


Whizlabs offers both educational resources and sample questions. The AWS Certified Machine Learning Specialty section contains one free practice exam of 15 questions. There are three full practice exams that cost around €20 each, but with €30 you can buy a 3-months standard subscription, which allows you to view all the courses available on the WhizLabs, plus their sample questions.

I used their practice tests to see if I’m ready for the real exam, and I’m glad I did so. On my first try, I scored 74%, which showed me that I’m not on top of the material, as I imagined I was. I mostly failed to correctly answer questions about specific configuration details, for SaaS and streaming services. I tried to convince myself that the Whizlabs’ practice exam was a lot harder than the real exam. A guess that proved to be wrong when I took the real one. In fact, I think the difficulty was a lot like the real exam, and some questions were even duplicates of what I had seen on Whizlabs.

I strongly recommend retaking the practice exams until you score 100% on all of them. Do this in intervals, so you have enough time to forget what you’re going to forget!


Pluralsight has a path of 7 courses dedicated to the exam. I already had a subscription, so it only made sense to go through the material, and to be fair, it is quite extensive and in general covers the topics. But then again, I would recommend the aforementioned Udemy course over this one.

A Cloud Guru

A Cloud Guru also has a dedicated preparation course for the exam. In my opinion, the courses offered by PluralSight and Udemy have a higher quality and better prepare you for the exam. But it comes with a full practice test and free hands-on labs in the AWS environment. However, I would only recommend buying a subscription if you are also interested in learning more cloud concepts.

Aside from the resources mentioned above, you can find many more on the internet. Obviously, I have not had the chance to try them all. There are also some free sample questions, but I have been dubious about some of the provided answers. You can, of course, use them with caution. However, after a point, the incremental time spent to find and study resources does not translate to a significantly higher exam score. There’s a point where enough is enough, and you need to take the plunge!

And with that, let’s get to how to register for the exam.

Exam Registration

Logging in

First, you must login to AWS Training & Certification website. You have two options for logging in:

Logging in with AWS Partner

If your current company is an AWS Partner, you can use your business email address for logging in. But note that once you acquire a certificate with this business email, you cannot change the e-mail address later easily.

If you leave your company, you must write to the Support team and inform them that you will no longer have access to your work email. Then they would help with changing your email address. But it’s better to avoid such complications if you have no definite reason for registering with your work email.

Logging in with Amazon

This is the same account you use for shopping from Amazon. It would be a valid login option.

At the same time, if your employer wants to benefit from having your certificate, you can authorise this in your dashboard and connect your account to your employer’s AWS Partner account.

When logged in, you will be directed to the certmetrics website.

Accommodation for non-native speakers

If your native language is not English, and you choose to take the exam in English, you can benefit from an extra 30 minutes.

The official exam duration is 180 minutes (3 hours), which in my opinion, is just right. It’s neither too long if you take your time reviewing questions nor too short that pressures you. But better safe than sorry. Although I didn’t myself, you can request a non-native accommodation in your dashboard.

However, it’s nice to know that the exam is also offered in Japanese, Korean, and Simplified Chinese.

Choosing the exam provider and its type

You have the liberty of choosing a delivery method from two different providers: Pearson VUE, and PSI. If one is better than the other, I wouldn’t know because, for both my certificates (ML Specialty and Cloud Practitioner), I used Pearson VUE, and I was mostly satisfied with my choice.

You also have the liberty of choosing on-site or online exams. I chose online, as I didn’t have a reason not to. However, if you don’t have a stable internet connection or a quiet and decluttered space to take the exam in, you might want to consider on-site exams. Unfortunately, they are mainly offered in big cities. In the Netherlands, where I live, the on-site Pearson VUE exam is offered at test centers in Amsterdam and Utrecht.

Saving costs

The ML Specialty Exam costs around $300 (excluding taxes), which many might consider as expensive. This is why many might want to know how they can get discounts.

As stated in the benefits section, you’ll get a 50% discount coupon each time you pass an exam. So if you, for example, take the Cloud Practitioner exam first (which costs around $100) and then take the ML Specialty, you would end up paying 100 + (300/2) = 250 for two certificates, whereas going straight for the ML Specialty will cost you $300 for one certificate. Plus, achieving the Cloud Practitioner certificate can help you prepare for the ML Specialty.

Checking in for the exam

If you have opted for online exams, make sure you have all the necessary requirements ready and tested. In Pearson VUE’s case, it involves running their application and testing your hardware devices. You can check in on your exam day 30 minutes prior to the start of the scheduled time, which I recommend you do, in case unexpected problems arise.

Have a valid ID at hand. Passports, residence permits, and driver’s licenses are acceptable. If you are originally from China, Cuba, Iran, North Korea, Sudan, and Slovenia, you would need an ID issued from another country, or they wouldn’t let you take the exam.

Make sure the room you are taking the exam in, is quiet, your desk is empty, and there are no electronic devices in your proximity (even smart watches). The proctor (your exam supervisor) would ask you to take pictures of the room. Also, make sure nobody comes into the room while you are taking the exam, otherwise, you would be disqualified. I specifically asked my roommate NOT to bring me coffee! :D

No breaks, even for going to the bathroom, are permitted during the exam. So for the 3 hours-long exam, I would strongly recommend you empty your bladder and not take much liquid before. However, you are allowed to bring a water bottle to the exam session.

Make sure you are well rested, as 3 hours staring at the monitor and thinking about details that might trick you is a tiring task.

During exam

Read each question carefully. Most of the questions come with a scenario. Pay attention to details that are intentionally put in the question body. Don’t get excited when you read the first few words. Chances are that there would come a phrase later on, that falsifies your first guess. Pay double attention to the last sentence. Look for phrases that say “most important”, “in the most efficient”, or “cheapest way”. These can be game changers.

Don’t spend too much time on one question. If you’re unsure about it, flag the question. When you’re done with the rest, you can spend the remaining time on the ones you have flagged. This is, in my opinion, the most efficient way.

Don’t get overwhelmed if you don’t know the answer to multiple questions in a row. It is supposed to be like that, and most probably, others feel the same way. Take a deep breath and remind yourself that you have read this in our blog!

The proctor will warn you if you look anywhere but at your screen (so you can’t stare at the ceiling while thinking about the question). Staring for 3 hours straight at a screen can take a toll on your eyes, so my suggestion is to shut your eyes while looking at the webcam from time to time. As stated before, you’re allowed to bring a water bottle to your exam session. Take sips from time to time. It helps keep you refreshed.

Exam results

Pass/Fail Criteria

According to the official AWS Exam Guide your score will be scaled in the range of 100-1000 and you would need to achieve at least 750 points to pass the exam. They use a difficulty scaling model to account for distribution of difficult questions in the version of exam you are taking.

When to expect results

Immediately after your exam, you’ll get the result in terms of whether or not you passed. If you did, you’d receive your certificate and report of your score a couple of days later.

The impact of studying for the exam

Aside from the certificate, you might be interested to know whether studying for the exam can help boost your knowledge. To be honest, I think the answer to this question is highly dependent on your current knowledge of data science and AWS. For me, as an AWS data scientist consultant who uses SageMaker on a daily basis, studying didn’t make much of a difference. Of course, I had to memorize, for example, what type of instance each built-in algorithm supports, but these things are usually a google-search away.

However, if you are new to AWS or SageMaker in particular, studying for the exam can provide a road map to get up to speed.

That’s it! If you have additional questions, or recommendations to make this post better, please reach out to me via email.

What is AWS ML Specialty certification?
The AWS Machine Learning Specialty certification is a professional-level certification that validates a candidate's knowledge and skills in developing, deploying, and maintaining machine learning (ML) models on the AWS Cloud.
What are the prerequisites for AWS ML Specialty certification?
Candidates should have a strong understanding of ML concepts and techniques, proficiency in at least one programming language, experience working with AWS services and architectures, and hands-on experience developing and deploying ML models on the AWS Cloud.
What topics are covered in the AWS ML Specialty certification exam?
The AWS ML Specialty certification exam covers a range of topics, including data engineering, exploratory data analysis, modeling, ML algorithms and techniques, evaluation, deployment and implementation, and AWS services for ML.
What are the benefits of achieving AWS ML Specialty certification?
Achieving AWS ML Specialty certification can help professionals to demonstrate their expertise in the field of machine learning and increase their credibility and marketability to potential employers. It can also provide access to exclusive AWS resources and networking opportunities and help professionals to stay up-to-date with the latest developments and best practices in the field