How I tackled the AWS Machine Learning Specialty exam (NEW — 2023)
Hi and welcome to this blog post. If you are reading this article, you are probably interested in doing the AWS ML Specialty exam. I took the exam in February 2023 and passed it with a score of 859 (750/1000 needed for pass); and in this article, I will try to give you some ideas (and plenty of links to the resources) on how to prepare for the exam and share my scores from the practice exams which you can use for benchmarking. If you are considering getting AWS ML certified and willing to invest time and energy for this purpose, definitely go for it. It is considered a challenging exam and it really is, but with proper preparation it can be tackled. And once you get the certificate, you will feel that it is worth the effort.
The outline of this post is as following:
1. My background
2. Resources for exam preparation
3. My approach to the exam preparation process
4. Conclusion and looking forward
Let’s get started!
My background
I have bachelor’s degree in Industrial Engineering and master’s degrees in Industrial Management and Supply Chain Management. Machine Learning was not the core focus of my studies, nevertheless, we had some exposure to ML and Data Mining concepts. I fell in love with the magic of math and numbers, so continued to develop myself with online courses and participated in the online hackathons. I have 8+ years of work experience in project management, chemicals manufacturing and logistics sphere. I have tried to apply ML concepts during these work experiences. For example, when working for one of the chemicals manufacturing company, I tried to use state-of-the-art methods to forecast the raw materials needed for production. They appeared to have much less error margins when compared to the actual method. It was not directly in my job description, but I try to be proactive and unleash the vast potential of the data that companies have. In my current role, I build simulation models of the processes and from time to time, I apply ML algorithms instead of using rule-based algorithms.
I am also one of those who believes in the strong math/statistics foundation for prospective Data Science (DS) and Machine Learning (ML) practitioners. It is true that the generative AI models and no-code/low-code software make our jobs easier and they do some part of the heavy lifting. But from my experience, the most difficult part is not the coding itself, rather knowing which DS/ML tool to use for answering the business questions. And at this stage, deeper knowledge of the algorithms’ under-the-hood working principles help me.
You don’t need to be proficient in all the algorithms; for example, my experience was more in numerical methods, I never had exposure to text classifications and sentiment detection models except for a few toy models. If you have never done an AWS certification, I suggest that you start with the AWS Cloud Practitioner exam. It is cheaper, it gives you experience on AWS exam logistics, it familiarizes you with different tools in AWS, not only the ones that are used for ML. I did the same — passed the Cloud Practitioner exam in September 2022. And it is financially better, because you will get 50% discount for your next exam.
How much time is needed for exam preparation? This is a highly subjective question, and the answer can vary from person to person. For me it was two months, studying only on the weekends.
Resources for exam preparation
In this section, I will explain the resources I used for exam prep and share their upsides and/or downsides. Please bear in mind that there are several other resources also which I did not have a look at all. Reviewing them also might be helpful (for example, the other AWS ML prep courses on Udemy by Kane and Maarek). Every person has his/her own learning style, so, please feel free to adjust what I explain here to your own style. The first thing I did was downloading and printing the official guide from AWS website, posting it to the wall of my room.
These were the resources I used for preparing for the AWS ML Specialty exam:
1. Udemy course by Chandra Lingam:
Only course that goes really deep into the topics. This course helped me to answer multiple questions in the exam. Their only drawback is that the practice test was slightly easier than the actual exam. Chandra teaches the best practices (spot training, for example), provides lots of example notebooks and reusable codes/functions that you can use for your personal and professional projects.
2. ACloudGuru:
This course is very good, but some parts are outdated. The course was recorded some two years ago, and during this time, some features and UI have changed. I liked the way they explained the concepts starting from their names. For example: OSB = Orthogonal Sparse Bigram. Orthogonal = independent, Sparse = because you skip some words in between, Bigram = two words.
3. WhizLabs:
Apart from the course, they have 3 practice tests which are more difficult than actual exam (which is nice!). These were my notes about this course:
- The instructor goes the long way sometimes.
- Some commands are outdated (like retrieving image uri).
- Explains only built-in algorithms.
- Questions are of high quality and close to actuals.
I actually found those outdated functions and methods useful in the sense that I had to search on the internet to find out the recent ones which resulted in better learning than just copying the codes from the instructor.
4. The book:
This book was written by AWS employees and they cover everything in the book. I found online courses more useful than reading a book. However, they provide you an access to online practice exam which is very close to the actual exam. Another good thing about the book is the flashcards which you can use to make sure that you know most of the concepts by heart.
Its resources on the website look like this:
You can do the exams in study or exam mode (exam model will be timed and it will tell you your score at the end). It has separate practice questions for each chapter in the book.
AWS SkillBuilder has many courses related to SageMaker and Machine Learning algorithms in general. They have a “basic math for ML” course which might be useful, if you have gaps in basic math/stat knowledge. You can also use this free online game-like tool (simulator) if you would like to better understand the Neural Networks, for example.
I do not mark this as a separate bullet point, but SageMaker developer guide and other help sections are of paramount importance. They will be the first place that you will look when things don’t go well in the Jupyter notebooks and when you get error messages regarding the algorithms.
I explained the courses, mock exams and books I used for preparing for the AWS ML exam; but before jumping to the next section, I would like to share with you my scores from those practice tests. As you can see, some of them are not that bright. But that is okay 😊
My approach to the exam preparation process
My (and many others’) first tenet is that there is no better way of learning than doing it yourself. I have seen some cases where one tries to pass the exam, memorizes the questions and their answers without hands-on experience. Believe me, it won’t have much impact on one’s career. Use the advantages of the free tier offered by AWS. If your free tier has expired, you can still use for example ml.t3.medium instances for your analyses. They costed me less than $10 per month. You will be paying $300 for the exam, and in this case, $10/month should not be a big deal. It is also important to know what actions in SageMaker costs money and what not. If you have not used SageMaker a lot, this might be a worrying topic when getting started. Endpoints cost money. Notebooks cost money when open and running. Stop the notebook once done and you will not incur any charges. Endpoint configurations don’t cost anything. Naturally API calls or direct usage of AI services (Transcribe, Polly, etc.) are charged. You can read all the cost details from AWS webpages.
My second tenet was treating the exam prep as a project. I created an Asana account and created six main topics on the board: Data Engineering, Exploratory Data Analysis (EDA), Modeling, Algorithms, ML Implementation, General and AI Services. Then I placed the tasks beneath and set deadlines for tasks. You can always update the tasks, deadlines, add new ones, etc. You can see a screenshot from my Asana below:
Nowadays any company has data. If you are working for a startup, it might be in Excel files, if you are working for SME’s, they might be using some databases or cheaper versions of Enterprise Resource Planning (ERP) software, or if you are working for large companies, they might be using SAP, etc. Think about how can you use the available data for preparing for the exam and also doing something useful for your company. It is a win-win situation. Use that data to apply XGBoost, Regression, or any of the classification models. I want to draw your attention to one important detail here: if your company has an AWS account, then it is awesome. If not, please be careful with data, get approval before moving company data to any other account.
I was using the mock exams to identify my weaknesses and adding them as a separate task to Asana. I was practicing them until I become proficient. For example, as I am not a software developer or computer engineer, my knowledge of securing SageMaker instances and networking topics were limited. I spared more time on these topics. Regarding the resources I mentioned in the previous section, you don’t need to watch every video and do all the exercises. Skip the parts that you are proficient in and watch the videos in 1.25x or 1.5x fast mode if you are very familiar with the content, but just need to refresh your knowledge.
Before and on the exam day
It is strongly suggested that you do the exam in-person at a test center for several reasons:
- You will get pen and paper to take notes, which you can’t do online.
- If for any reason the internet disconnects, it will be the test center’s responsibility. But if it happens at home, you might end up with fail and no refund.
Make sure to visit the test center and familiarize yourself with the location before the exam date.
Sleep the night well before the exam without any distractions. Try not to open social media, etc. not to fill in your brain with information that you don’t need (at least for today). Have a good breakfast and come to the exam location 30 min earlier. In my case, bringing water bottle to the desk was not allowed; make sure you are hydrated before entering the exam room. Nevertheless, invigilators let me drink water during the exam by getting up and leaving the classroom. After I drank water, they let me in again. Please ask your local test center about water and toilet needs. I do not know if the rules are universal or not (logically they should be).
Needless to say, that you can mark the questions that you are not sure and come back later. In the first round, I did not think on a question for more than 1–1.5 min, if I can’t find the answer, I was selecting the answer that I thought to be the best, marked the question and moved forward. Once you are done with the first round, you can come back to those difficult questions more relaxed. Wrong answers are not penalized, don’t leave any question blank. If you randomly answer 4 questions, 1 of them should be correct, at least according to the probability theory. 😊
After the exam, take a walk, sit back and relax. Your result will not be shown on the screen and instead mailed to you within 24 hours (although official page says it can take up to 3 days). Try not to think a lot about the exam. Past is past and worries are not helpful 😊 Even though I passed the exam with a high score, after the exam ended, I was not sure if I will be able to pass the exam. So, it is natural if you have the same feelings.
Conclusion and looking forward
After the certification exam the journey does not end, on the contrary, you start a new path. Continue engaging with the community, become an AWS Community Hero, participate in local meet-ups, continue networking. These are the most value-added things for one’s career in the cloud. And probably you will also learn new topics and technologies that you were not aware of before. Try to apply them, gain hands-on skills and add them to your toolbox. You never know at which stage of your career they may help you solve business problems.
All the best with your endeavors for getting AWS ML certified!
If you found this useful, please follow me for more and consider giving applauds 😊 If you have questions or if you would like me to explain other topics, please feel free to post them as comment. I will do my best to answer them. And finally, let’s stay in touch! I am happy to connect on LinkedIn.
My LinkedIn profile: Yashar Ahmadov
