The AWS Machine Learning Engineer Associate is one of the two new beta exams announced recently. I share my experience preparing for and taking the exam.

1. Background

AWS recently announced two new certification exams: the AI Practitioner Foundational and the Machine Learning Engineer Associate. Both exams are in beta as of this writing, allowing AWS to collect feedback and performance data before their full release.

After taking the AI Practitioner exam a few days ago, I decided to round out my experience by scheduling the ML Engineer Associate exam as well. I noticed several similarities between the two exams.

For context, I passed the Machine Learning Specialty exam a few weeks ago. This certification covers machine learning and AI concepts at a more advanced level. It’s important to note that thorough preparation is key for any exam, including this one.

2. Preparation

I referred to several sections from the Enhanced Exam Prep Plan: AWS Machine Learning Associate (MLA-C01) course on AWS Skill Builder.

The lab exercises were helpful, though doing them alone is not enough to guarantee success in the exam. The data preparation, ingestion, and transformation domain is a major focus. If you don't have a strong background in data, it’s worth getting hands-on experience with services like Glue or Athena.

3. The exam

Overall, I found the exam appropriately challenging for the associate level. It was neither easier nor harder than expected.

3.1. Concepts and services

As anticipated, the exam was practical, with many questions focusing on configurations and settings.

The primary focus of the exam was on SageMaker. Most questions revolved around this service, covering data ingestion, preparation, transformation, bias detection, and mitigation. I also encountered questions on key steps in model performance evaluation and monitoring. It’s important to be familiar with tools such as Data Wrangler, Model Registry, Clarify, and Model Monitor. Expect questions testing your understanding of settings and configurations, emphasizing hands-on knowledge.

Candidates should also understand model deployment and inference options, including when to use each. Other required SageMake services include JumpStart, Autopilot, and Canvas. Essentially, any aspect of SageMaker could be tested.

My exam also featured some theoretical questions, such as precision, recall, accuracy and other metrics. You might be asked to consider how changing a term in a fraction formula impacts the result! While no direct calculations were required, these questions can take longer to process, especially if they come up unexpectedly in the exam. I also encountered questions about algorithms like random forest, XGBoost, and linear learner.

Generative AI topics were also present, though fewer than in the AI Practitioner exam. Expect questions on foundation models, large language models, RAG (retrieval-augmented generation), and Bedrock.

I also received questions related to standalone AI services.

The data domain is a major focus in the ML Engineer Associate exam. You should be familiar with services such as Glue, Glue DataBrew, Lake Formation, Redshift, and, of course, S3, which was covered in some depth, particularly around its features and configurations.

As with all AWS certification exams, security plays a significant role, and this exam is no different. Be comfortable configuring VPC security and using services like IAM and Macie.

3.2. New question types? Yes!

I was pleased to see the new question types: ordering, matching, and case study. These accounted for roughly 10% of the exam. The case study scenarios involved a series of 4 or 5 related questions. I appreciated the opportunity to dive deeper into specific scenarios without constantly switching the context, allowing for a more thorough exploration from different perspectives.

3.3. After the exam

The results, along with the two badges (one for early adoption and the standard certification badge), arrived quickly — about three hours after I completed the exam.

4. Comparisons

Here’s my subjective comparison of the ML Engineer Associate exam to other ML/AI certifications.

4.1. ML Engineer Associate vs AI Practitioner Foundational

I took these two exams just a few days apart, so both are fresh in my mind.

I found them to be of nearly equal difficulty. While the ML Engineer Associate exam wasn’t easy, I found the AI Practitioner exam to be more challenging than I would expect from a foundational-level test.

The AI Practitioner certification focuses heavily on generative AI and related concepts like responsible AI, explainability, and interpretability. In contrast, the ML Engineer exam is around creating, deploying, and operating ML pipelines, with a strong emphasis on SageMaker. The ML Engineer exam also contains more theoretical questions.

As with other AWS certification pathways, if you pass the ML Engineer Associate exam after obtaining the AI Practitioner certification, you’ll be recertified for the foundational certification. The new expiration date for the AI Practitioner will match the ML Engineer certification.

4.2. ML Engineer Associate vs ML Specialty

The ML Engineer exam is positioned to be more practical, while the ML Specialty exam focuses more on theory. Both exams heavily feature SageMaker, and in my view, they are more closely related to each other than either is to the AI Practitioner exam.

The two exams cover similar services and concepts. The ML Engineer exam delves deeper into service settings and configurations (such as SageMaker features and hyperparameter tuning), while the ML Specialty exam leans more toward metrics and algorithms.

I don’t think the difference between the two exams is significant, so I’m curious to see what future career awaits the ML Specialty certification. 😉

5. Summary

Overall, the Machine Learning Engineer exam met my expectations for an associate-level certification. The exam is practical, with a strong focus on Amazon SageMaker.

I highly recommend getting hands-on experience with SageMaker and relevant data services before attempting the exam. Relying solely on reading or watching video courses won’t be enough unless you can memorize the entire documentation. Practical experience is key to success in this exam.

6. Further reading

AWS Certified Machine Learning Engineer - Associate - The exam's page on the Training and Certification website

AWS Certified Machine Learning Engineer - Associate (MLA-C01) Exam Guide - The exam guide