Passing the AWS Certified Machine Learning Specialty Certification (2024)

“Congratulations on passing your AWS Certification exam!” This is one of the most satisfying messages you can see in your inbox after embarking on a journey to get a new cloud certification. The AWS Certified Machine Learning Specialty certification is a great way to showcase that you know your stuff when it comes to end-to-end machine learning in the cloud. It covers topics such as:

• Engineering and preprocessing data before it even touches your machine learning models
• Building, assessing, and maintaining models
• Productionalizing models with MLOps
• And surprise, surprise, knowing all of the major AWS Machine Learning services and capabilities

While the certification alone can show that you are able to apply ML in an AWS environment, it is really targeted at people with previous ML and AWS experience. AWS recommends that you have 2+ years of experience in developing, architecting, and running ML or deep learning workloads in AWS Cloud. If you have experience with ML but no AWS experience consider looking into the AWS Cloud Practitioner certification to get accustomed to AWS services and/or experimenting with AWS ML services before going for this certification. If you have no ML experience at all, I recommend looking into other ways to build up your ML knowledge first.

This certification is a 65-question 3-hour long multiple choice test. It took me 5 weeks to study and I spent around 15 hours each week. Below are some of the resources I found helpful.

Frank Kane’s AWS Certified Machine Learning Specialty Course

Frank Kane/Stephane Maarek’s courses are always my first step in studying for a new AWS certification. If you can understand all of the topics and services in this course, you have the foundational knowledge needed to pass the test. This course covers everything you need to know and provides some hands-on labs that shows you some of the services in action. Make sure you take extensive notes if you take this course, as this makes for easier review later on. The Udemy course can be found here. If you’re paying out of pocket (my company luckily has a Udemy Business subscription) then be sure to look for discounts on Datacumulus or during Udemy sales.

SageMaker Documentation

While it may seem like overkill, I found that reading through the SageMaker documentation helped solidify my understanding of some of Frank Kane’s content. SageMaker is AWS’s top ML service and is covered extensively on the test. I found it most helpful to read through after completing Frank Kane’s course so I could know which parts of the documentation to delve into more and which parts I could skim through. I focused mostly on sections about data preparation/processing and model training, deploying, monitoring, and evaluation. Overall this took me 4–5 hours total to go through over the course of a couple nights. The documentation can be found here.

Practice Tests & Questions

After making sure you have a foundational understanding of ML and AWS, going through practice questions is the best way to prepare for the exam. Online courses and documentation don’t fully help you understand how to apply the topics and practice questions help to synthesize everything together. Below are some of the questions sets I found helpful:

• Official AWS questions: These free questions are the most true to the real test so make sure you take advantage of them. Between this PDF and the questions found in your aws.training portal, there are around 40 unique questions (some of the questions are recycled).
• Questions at the end of Frank Kane’s ML course: The course only came with 10 questions, unfortunately. He also has a separate course with the full practice test here but it was not offered on my company’s Udemy Business account so I did not take it.
• Abhishek Singh offers 2 full-length practice exams with a third 10-question warm-up test on Udemy. These were the only full-length tests that I used so I found them very helpful. This course can be found here.
• Exam Topics has a decent set of questions. The first 80 are free. One call out is that the option(s) labeled as the “correct” answer was sometimes different than the most-voted answer. I found the most-voted answer to be the correct answer in most cases and was able to validate it via the comments and ChatGPT. Side note: ChatGPT was a must-have in my training to help me understand practice questions better.
• Tutorials Dojo offers 10 free questions and some paid practice tests. I only went for the free questions.
• Whizlabs offers 15 free questions and some paid practice tests. I only went for the free questions here too.

Check Out this Article

This Medium article by Collin Smith was my go-to blog while studying. The tips at the end were great to know going into the test. I made sure to re-read this article the night before my test day.

Tips

Below are some tips I put together based on my experience with studying for this exam:

• Read the questions carefully. Understand the problem statement and the constraints. Many questions have potential answers that would solve the problem but the constraints (e.g., costs, operational overhead, experience level of a company with data engineering and machine learning) made some more correct than others.
• Take your practice tests in an environment similar to the actual test. That means no phones and no distractions. This will help make sure you have the endurance to get through all 65 questions on game day.
• As you’re going through practice questions, document any time you’re uncertain of yourself. Follow up afterward to fully understand what the question is asking and why the correct answer is correct. If you get the question wrong, make sure you understand why your answer was wrong. Take notes on all of this — they will create a good personalized study guide. Documenting whenever you have an understanding gap provides you with a great list of topics to make sure you go into more depth as you study.
• Find good opportunities to multitask and review your notes. While this might not be for everyone, I reviewed my notes on the stationary bike or while walking on the treadmill. Reviewing my notes multiple times throughout my preparation really helped to lock in the information I needed to know.
• Pair ChatGPT with your practice questions to delve deep into the topics they cover. I generally found low hallucination rates while using ChatGPT 3.5 to help me understand test questions better.
• In the 2 days leading up to the exam, retake some of your past practice tests, re-read the medium article I posted above, and go over all your notes. The day before the exam, retake the AWS-provided question sets as these are the closest examples to what will be on the test. Make sure you go over all of your previous understanding gaps to ensure you’re good to go in those areas.
• Make sure your mind is fresh going into the test. Don’t grind too late the night before. Studying right up to when you sleep is a surefire way to feel burnt out the next morning. I woke up early, got a workout in, and had some coffee before my 9 AM test to clear any mental fog. Do what will work for you!
• Be consistent with your studying. Being regimented will help you get the most out of your time and help you retain information better.
• As always, set a date for your exam early into your studying. Having a deadline forces you to be on top of your studies. Plus knowing when you’ll be done with the grind is a benefit.

About Me

I graduated from the University of Virginia with a Systems Engineering degree in 2022 and took a ton of courses on ML/AI. I also have some experience outside of the classroom bringing machine learning solutions to a few different companies and organizations (one of which used SageMaker). I currently work as a technology consultant at Pariveda Solutions in a Cloud Engineer role. In the past 6 months, I have been in a DevOps role for my current client and I am hoping to use this certification as a stepping stone into MLOps and general AI/ML in the cloud.

My company recently hosted a hackathon for social good where I worked with a nonprofit client to deliver a machine learning framework that predicts county-level adverse mental health outcomes. This experience reinvigorated my desire to dive back into AI/ML and motivated me to go for this certification. I’m super excited to continue my learning journey in this domain, as AI/ML is the present and the future. Stay tuned!