Passed my AWS Solutions Architect Exam!

So after talking about getting the AWS SAA-CO2 certification for about a year, I have finally taken the test, passed and obtained my certificate.
Having done my last ever test about five years ago in university, I have to admit it was quite difficult to get back into the study mindset and set goals for yourself when you don’t have a semester timetable breathing down your neck.
The test was definetely not easy and initially I thought I was over-prepared, now looking back I am certainly glad I took the time to study, take practice tests and consult colleagues who were experts in AWS.

Why I got the certification

So the reasons I wanted to obtain this certification:

• I wanted to be self-reliant when using the cloud on a basic level.
  
• I believe this adds to my Career Capital [1].
  
• I wanted to be able to be a full-stack, end to end Data Scientist.
  
• I found the control and multitude of cloud services a liberating experience.
  
• I wanted to focus on the core instead of always having to worry about context [2].
  

I believe that no matter what kind of Data Science work you do, unless you’re already in a large tech company or FAANGs, there will be some aspect in your work that will involve the cloud in some form.

In order to be an effective Data Scientist it is therefore necessary that you understand the basic ideas on how to utilise the cloud. This will not only help you work together with your Data Engineers and UI/UX developers but also help you be a more valuable member of your team.

There’s nothing worse IMO than thinking you can just write a giant notebook file (maybe if you work in Netflix) and chuck it over to the Data Engineers to worry about productionising your spaghetti mess of a code. Or finding out you can’t simply train your model on a laptop anymore and now need a bigger/faster compute without submitting your hardware requistion form and waiting for half a year to get a reply.

Some of you might think that these sound like a Data Engineer’s job, but my take is that unless you’re in some very specialised ML/DL research position where you really are spending all day with algorithms and math, it really should be partly your job as a Data Scientist to do some of these work as well. Otherwise you’re job is not too different from a Kaggle Notebook (nothing wrong with Kaggle) and I think would one day be easily replaced with Auto-ML solutions (have a look at this if you don’t think this will happen eventually).

Resources I used to prepare

There are three main resources I used to prepare:

1. Adrian Cantril’s AWS course:
   Best $50 I ever spent IMO. There’s a reason his course is so highly recommended on the AWS subreddit when people ask for resources to learn AWS. Great content and fantastic explanations using diagrams and examples.
   

2. Jon Bonso’s Practice Exams:
   Second best $30 I spent. I don’t think I would of passed the test without having done this practice tests. The practice test that AWS provides or sells is simply not enough IMO. Some of the questions on the actual test were nearly identical to Jon Bonso’s tests.
   

3. Alozano’s AWS course notes:
   Best $0 I spent. His notes are based on Cantril’s course and I used it to cross check against my notes to ensure I didn’t miss anything.
   

Strategies

So all together it took my about four months to study for the test.
Some of the strategies I found that really worked for me were the following:

• Habit: I would start with a small block of time everyday where I would either watch a course video or revise some notes, this time block eventually increased from 30 mins to 2 hours and got to the stage where if I didn’t spend some time on the course I would get some nagging feeling in my head.
  
• Active Recall: I was watching through one of Cal Newport’s video where he went over the concept of Active Recall. This basically means to explain the material you just read out-loud as if you were trying to teach someone this concept. I initially thought this would be quite silly to be talking to myself but there was definetely a benefit to doing this.
  
• Spaced Repetition: This is basically Anki method where you would write up some flash cards with questions and test yourself on a repeating schedule where questions you got wrong would be repeated in a higher frequency and lower frequency for easy questions.
  Personally I used a free plugin for Roam Research since that was where I kept my notes in.

Taking the test

I chose Pearon-Vue to take the test. Using my Intel Macbook Pro 16”, I didn’t encounter any technical issues during the pre-test system tests.
They do ask you to take photos of the environment you’re taking the test in and it has to be clear of any unnecessary materials and additional monitors. I forgot about this just before the test and had to clumsily move my 49” ultrawide monitor to another table.

They also ask you to not move your face away from the camera or have anyone step into the room so make sure you tell people that you live with to leave you alone for the duration of the test.

The test was 140 minutes in length and I used up 90 minutes answering the questions and a further 30 minutes to review all the questions. I would definetely recommend reviewing questions or flagging them and coming back to them since I found a few errors in my initial answers which I corrected during my review.

Once you complete and submit the test you get your results straight away. You either pass or fail.

The actualy electronic certificate took about 2 hours before I got it from Credly.com.

What’s next

For the short term, I will be going back to reading and working on Data Science models and techniques (really excited about learning Graphical Models using Julia).
I am definetely planning to obtain another cloud ceritifcation this year, most likely the Azure Fundamentals (AZ-900) first then the AWS Machine Learning Specialty (MLS-C01).

---

References:

[1] So Good They Can’t Ignore You, Cal Newport.

[2] The Phoenix Project, Gene Kim.