Passing the AWS Machine Learning Specialty Certification Exam
I passed the AWS Machine Learning exam sometime in August and I’d love to share my experience and knowledge gained from taking the exam, to help anyone who’s preparing to take it or thinking of taking it.
I’d like to start off by saying that this exam is not a ride in the park as may be assumed, it has quite a high difficulty level. However, it is definitely not impossible to pass the exam, with efficient preparation, in the first sitting.
The Hows, What’s and Why’s
Why do you want to take the exam?
I believe the very first thing that will propel one to succeed in this exam is motivation or one’s level of motivation. I had scheduled to take the exam in mid-2020 — during the COVID lockdowns. Honestly, during that season it became almost impossible for me to prepare adequately for the exam due to various reasons. I couldn’t motivate myself to study for it. Also, there was no real personal motivation to take the exam except that, it was something I had included in my personal development plans for that year. I had just started using AWS as a data scientist and thought that certification will be a great milestone to cap off my experience using it, for the year. I ended up postponing taking the exam until last August. By that time, my motivation had become a pressing need to validate my ~3 years of experience in experimenting and operating machine learning in AWS cloud.
Having a solid motivation — whatever that might be — that drives discipline to study adequately for the exam, is definitely a good place to begin.
What do you need to do before beginning to study?
One thing to figure out is — what to expect from the exam, to be able to study efficiently and not waste time on unnecessary content. AWS has provided an exam guide (https://d1.awsstatic.com/training-and-certification/docs-ml/AWS-Certified-Machine-Learning-Specialty_Exam-Guide.pdf) to help one understand what domains and what type of questions to expect in the exam. The domains in the exam are:
- Data Engineering — 20%
- Exploratory data analysis — 24%
- Modeling — 36%
- Machine Learning Implementation and Operations — 20%
I believe another important thing to do is to book the exam date in advance. I’d suggest doing this after you’ve assessed your readiness for the exam so that you book a date that allows for enough time to prepare for the exam.
The exam costs 300 USD so you don’t want to pay that twice because you just didn’t have enough time to prepare. However, it is possible to cancel or reschedule your exam up to 24 hours before your scheduled appointment without additional fees. In AWS’ Certification FAQ (https://aws.amazon.com/certification/faqs/) they also say, “Please note that you may only reschedule two times after scheduling your original appointment. If you wish to reschedule a third time, you must cancel and schedule a new appointment”. Scheduling the exam in advance is a motivation in itself to find the discipline to study for the exam. I’d also suggest scheduling the exam not too far off from when you’ve finished preparing for it, so you don’t forget the intrinsic details in some of the domain content.
How to prepare for the exam?
Of course, everyone will prepare differently for the exam. So here I’ll just be pointing out things that really helped me.
Practical experience
First off, I think having a machine-learning background and some experience with AWS (especially using AWS ML services) makes the exam easier to pass. That means if you are lacking in any of those, you will need more time to study or need to study more materials than someone who has practical experience. The exam is very scenario-based - so it tests one’s practical knowledge and ability to solve various ML use cases and how to apply different AWS services to these use cases, rather than just theoretical knowledge about algorithms, for example.
Sample tests and practice exams
Doing many sample tests and practice exams really helped me get a grasp of what kinds of questions to expect in the exam, but also the skill needed to answer them. For example, a question may be quite long but there will be a few keywords present in the question which is your hint to the answer to that question. Practice exams and tests also helped point to specific topics that I still lacked knowledge of. This test site in particular was very useful for this purpose — https://www.testpreptraining.com/aws-machine-learning-specialty-questions
Courses
As useful as practice tests and exams are, they just won’t cover important details regarding the different domains in the exam. This is where it becomes handy to take a course or several courses in preparation for the exam. Many of the available preparatory courses on the internet for this exam are video lessons based. If you’d like to follow a course that’s not video-based and presents content according to the exam’s format, this site is probably what you’re looking for https://www.mlexam.com/.
Other video-based courses that were really helpful are —
- A Cloud Guru’s ML Specialty (https://acloudguru.com/course/aws-certified-machine-learning-specialty)
- Frank Kane and Stephane Maarek’s prep course on Udemy (https://www.udemy.com/course/aws-machine-learning/)
AWS SageMaker Documentation
Of all the tips I’ve written so far on how to prepare for the exam, I feel this is the most important. Why? Because there will be a couple of questions in the exam that have answers only found in the SageMaker documentation. My suggestion is to focus mainly on the training section of this documentation — https://docs.aws.amazon.com/sagemaker/latest/dg/train-model.html as these questions will mainly be in the Modeling domain of the exam. However, it could be broader than that. A review of this documentation a few days before the exam will come in very handy. Also, pay attention to panels within each page that have the subheadings ‘Important’ and ‘Note’. For example;
Exam Day Tips
- You have 3 hours for the exam. However, you’d almost definitely get through all 65 questions within 2 hours. Don’t rush to be done with the exam even if that means exhausting all 3 hours. Read every question multiple times, and engage and re-engage the questions as there are quite some tricky questions and options. Take note of keywords, and use the process of elimination as much as possible to arrive at answers to questions you are really not sure of.
- On the exam, you can mark questions you’re not sure of to come back to them. Don’t waste much time on a question that may be difficult to answer. Mark it and move on to the next question. You’ll have enough time to come back to negotiate it in depth.
