Anthony WatIntroduction
AWS has recently revamped their certification lineup to align with the growing AI/ML trend. Among them is the AWS Certified AI Practitioner (AI1-C01) exam, which is currently in beta. Beta exams validate exam questions to finalize the content before wide release. As an early adopter, you get a discount on the exam fee in exchange for your contribution to test out the exam for AWS. Given that this is an intriguing proposition and it aligns with my focus on AWS, I decided to write this exam and share my experience with the community.
A Bit About My Background
For context, I've been working exclusively with AWS for about three years and completed a couple of professional/specialty level exams, so I am quite familiar with AWS in general. My work also involves generative and conversational AI (although it has mostly been on the Microsoft side), so I have knowledge on concepts such as LLM and RAG.
I am interested in generative AI on AWS, so I've been experimenting on my own and has written a few blog post about Agents for Amazon Bedrock and Knowledge Bases for Amazon Bedrock. However, my knowledge of Amazon SageMaker is limited to understanding the ML workflow and some hands-on experience from AWS Workshops.
Overall I'd say that I know a bit more about AI/ML than the average Joe, so I was able to expedite my study somewhat. As you read about my exam prep, consider your own knowledge and experience to adjust your approach.
How I Studied for the Exam
In the past, I've always used A Cloud Guru to study for AWS and Azure exams. Recently I've been given an AWS Skill Builder subscription by my company, so I've decided to use it as the primary source of study material.
The official AWS Certified AI Practitioner webpage recommends a 4-step exam prep plan which I followed over two days. To test my knowledge prior to studying, I went through the official practice question set and got 85% which boosted my confidence (but eventually turned out to be a trap). I then proceeded with the enhanced exam prep course that included bonus practice questions and flashcards over the standard (free) version.
The course itself is well-organized and touches upon all topics. The instructor spoke very slowly in the video, so I watched the videos at 2x speed for a more reasonable pace. However, I found that certain concepts aren't explained very well and the level of details isn't representative of what the exam itself demands. I suppose the additional resources would have been good supplements, however I would prefer if the course itself is more in-depth.
The bonus questions provided additional practice opportunities and the flashcards were helpful for memorizing key concepts. I transferred the flashcards into a Word document for offline review before the exam. That being said, they did not cover all important concepts and I had to supplement them with my own research on Google (such as a list of ML algorithms).
Following my typical study regimen, I sought more practice right before the exam. Aside from the practice questions from AWS, I quickly went through Tutorial Dojo's free practice exams sampler given that they don't yet have a full practice exam package yet. (I later found out that Stephane Maarek does.)
How the Exam Went
I took the exam online at the comfort of my home in a quiet evening. The check-in process went smoothly and took about 10 minutes. The exam questions were more difficult than I expected of a foundational-level exam. On a scale of 1 to 10, I’d rate it a 4 in difficulty.
Although the exam guide lists case study as a question type, there wasn't any for me so your mileage may vary. There were many questions on machine learning concepts such as algorithms and performance metrics. Amazon Bedrock also featured a lot. There was also the expected mix of questions on Amazon SageMaker features, generative AI, and responsible AI. While I wouldn't say that the questions were very different from the official practice question set or the bonus questions in the exam prep enhanced course, they seemed more in-depth and specific.
It took me about 75 minutes to complete the exam, with the first pass completed in about an hour which left 27 out of 85 questions flagged for review. This is more questions than I usually flag in AWS exams (including the DOP exam), and after the review I was only confident on my final answer to about half of those questions. I finished the exam feeling I would pass, though not with full confidence.
A Retrospective on My Approach
I received the results by early morning and scored only 729, which admittedly is lower than I expected. Nonetheless, a pass is a pass. According to the report, I didn't do as well in the "guidelines for responsible AI" domain which was a bit surprising as I did well during practice. In any case, here is my reflection on my experience based on a typical retrospective framework.
What Went Well
Setting an exam date motivated me to be disciplined and keep to a set study schedule amid other priorities in life. The long weekend afforded me enough time to study, socialize, and do chores.
Following the official exam prep plan and the exam prep course helped structure my study. I've always studied by following a course be it from A Cloud Guru or AWS Skill Builder, and it's been proven to be a sound strategy.
What Didn't Go Well
I was overly confident of my hands-on experience with services such as Amazon Bedrock, when there were numerous features and other services that I've not have enough exposure to. Consequently I did not allocate time to watch tutorial videos or do hands-on labs to gain the necessary familiarity and "muscle memory".
I didn’t spend enough time on Step 2 of the 4-step plan to explore additional courses outside the AWS Skill Builder exam prep course. I've retroactively looked at some of the recommended courses and they would have improved my overall knowledge for this exam.
What Could Be Improved
Although I was able to get by with one full day of study, it would have been better to spread the study across multiple days for a better pace.
I would have done better with additional research on general AI/ML concepts such as algorithms and performance metrics, which the exam seemed to have more focus on.
Although the AWS Skill Builder exam prep enhanced course was decent, it was not fully adequate as the sole study material. Investing in additional courses and practice exams, such as those from Stephane Maarek, would probably have helped boost my score.
Summary
I hope this blog post gives you a sense of what to expect from the AWS Certified AI Practitioner (AI1-C01) Beta Exam. It's certainly not an exam that you can wing, unless you are well-exposed to AI/ML or are already studying for other certifications such as AWS Certified Machine Learning Engineer - Associate. However with the right material and a few days of study, it is very much achievable.
Check out the Avangards Blog for more articles on AWS, Terraform, and other topics. Best of luck with your studies, and I hope you’ll soon be certified!
