Passing the Beta AWS ML Engineer Associate and AI Practitioner Certifications
My experience with the new AWS ML beta exams and some tips
Published Oct 28, 2024
AWS announced the release of two exams AWS Certified AI Practitioner and AWS Certified Machine Learning Engineer – Associate, early August 2024 in this aws blog. The AI Practitioner beta exam was available to take between 13 Aug 2024 and 7 Oct 2024 and the MLE Associate beta exam till 28 Oct 2024 .
The AWS Certified AI Practitioner exam, as per the exam guide, focuses on AI, ML, and generative AI topics with specific emphasis on AWS specific services including Bedrock and Amazon Q from a GenAI perspective, as well as more traditional ML service. The ML Engineer Associate exam as described in the exam guide focusses mainly on using AWS services like Sagemaker and others to build, operationalize, deploy, and maintain machine learning solutions and pipelines. This is more focused compared to the ML Speciality Certification which is a lot broader and in addition tests on general Machine Learning theoretical knowledge as well as having deeper knowledge on wider range of AWS services for carrying out data engineering tasks.
This was my first experience taking a beta exam so i was not sure what to expect. I had missed the chance to take the Data Engineer Associate Beta last year so I was determined to squeeze these two into my schedule. Here are the high level benefits and downsides to taking the beta exam, which i have summarised from my experience and the aws documentation
Benefits
- The exam price is also 1/2 price compared to standard exam price. Note that if you have a 50% benefit discount voucher (from passing a previous AWS certification), you could also apply it when registering and get a further discount on the exam booking fee !!
- You also get an early adopter badge which you only get for taking the exam. You can achieve this if you take it before a deadline from when it is released as detailed in this blog
- Having been the first bunch of candidates to have taken the exam when it was released, you have bragging rights !! Your feedback and exam results help to shape how the AWS certification team shapes the final certification when the beta period ends.
Cons
- The exam has 85 questions instead of usual 65, within the same time period of 170 minutes, so you have less time per question and need to time yourself accordingly.
- Not a wide range of practice material available as these exams are in early release/testing phase.
- Questions and answers in beta exams may be unclear and may take a while to comprehend and answer. The AWS certification team usually fix these when the final version is released after the beta phase completes based on performance and feedback from candidates.
- Not all questions are graded. You will not know which questions are graded and which are not.
- You cannot take a retake in the beta phase if you fail the beta exam. You will have to wait till the final/standard version of the exam is released after the beta phase.
For preparing for the exams, I arranged them at least couple of weeks apart and close to the respective deadlines to give myself enough time for preparation due to other commitments. I was also due to do my renewal for the ML speciality Certification before the first week of October so I made sure I got that out of the way first before attempting the beta exams, as there was a good deal of overlap in content.
The main issue with beta exams is that there is not a lot of material/courses for preparation. I sued the following and would follow the same "tactic" if I had to do a future beta certification.
- AWS Skills Builder for the exam prep and practice sample questions. Every new exam, should have AWS skills builder course and sample questions available for free. For the enhanced verisons, you will need to have a paid subscription to Skills Builder. For the AI Practitioner, you can access them here and for the MLE Associate here
- Data Engineering services like Glue, Redshift, Kinesis (relevant AWS documentation)
- Machine Learning Lens Whitepaper
- Tutorials Dojo Practice Questions for AI Practioner and MLE Associate (Paid)
- Some practice on my AWS account using any features I was not familiar with
After 3-4 weeks of study/practice using the resources above, I did both exams in a local testing centre on separate weekends. The passing score for AI Foundational is 700 whilst the Machine Learning Engineer Associate is 720. I got the results via email, a couple of hours after I completed the exam but this may vary for different people taking the exam.
AWS has also recently announced introducing a new style of exam questions to include ordering, matching and case study question types in addition to standard multiple choice questions.
These had not featured in any of the exams I had taken upto early October 2024. But these did feature in the AWS ML Engineer Associate Beta exam i took on October 27th 2024. So they will possibly start featuring in all exams going forward. My personal feedback on this was positive and I felt it broke the monotony of standard multiple choice questions.
An example of ordering and matching questions for MLE Associate and AI Practitioner exam sample questions from Skills Builder is shown below
The case study question especially gives you one case study and then 4 follow up shorter questions based on the same case study. This meant i could finish 4 questions a lot faster as I did not have to read a brand new unrelated question every time. In the end, I did finish with a lot more time to spare than I normally would have had. In the past I have found a lot of the questions quite lengthy and having to re-read them again (mainly for the speciality exam) and have almost run out of time.
An example case study question from the AWS Skills Builder Sample Questions is shown below
From my experience, I found both exams were challenging, even though I have an ML background and experience with AWS ML services. Do not take the AI Foundation exam for granted, as it is harder than a typical "Foundation" exam. For context, I scored a lot lower in this exam, compared to the ML speciality and MLE Associate !
The Associate MLE Exam in comparison to the ML Speciality, focussed a lot more on using Sagemaker features (Feature Store, Model Training, Model Monitor, Deployment options, Security etc) across the ML lifecycle as detailed in the Machine Learning Lens whitepaper and a lot less on theoretical ML aspects (the ML speciality in contrast tests on both). It will also test you on batch and streaming data preparation skills using services like Kinesis and Glue/Glue DataBrew.
On a scale of 1-10 in terms of difficulty, if the ML speciality exam had a difficulty of 8, in comparison I would rate the AI Practitioner and Associate MLE, 3 and 6 respectively. But this is purely my opinion :)
Passing the AWS Certified MLE Associate Exam, also automatically renews the Certified AI Practitioner Exam if you already hold it. I was not aware of this, until i got an automated email regarding "recertification" of AI Practitioner once I passed my MLE Associate Exam. So in theory, in the future, you only need to renew the MLE associate exam to renew both.
I am still not clear if renewing the ML Speciality exam would renew both the MLE Associate and AI Foundation exams in the future. I guess I will only find out in 3 years :)
Luckily after I got the good news about my results, I am now the proud owner of these two beauties. Hopefully, you can also do the same when another beta exam gets released in the future !
