Experience of the AWS Certified Machine Learning Engineer — Associate training and beta exam
AWS recently released two new exams, the AWS Certified AI Practitioner and the AWS Certified Machine Learning Engineer — Associate. This post covers the Machine Learning Engineer — Associate training and beta exam after I recently studied for and passed the exam whilst it was in beta.
The exam is marketed as being at associate level. However either I’m out of touch with what it takes to pass an associate level course these days (I’ve not sat an original Develop, SysOps or SA Associate since 2017 as I’ve renewed them taking the relevant pro exams) or the bar is significantly higher for the newer associate exams, as I felt the same way when I did the Data Engineer Associate when that was released.
In keeping with AWS’ approach to the other exams covering AI / ML (AWS Certified AI Practitioner and AWS Certified Machine Learning — Specialty) the exam doesn’t just test your knowledge of AWS services, it also tests your general knowledge of machine learning and artificial intelligence theory. The end to end lifecycle is tested, so an understanding of the data ingestion and preparation phases and the associated AWS services is important too, in this area there’s plenty of overlap with the content tested in the AWS Certified Data Engineer — Associate exam. The full exam guide provides the full list of areas the exam tests against — https://d1.awsstatic.com/training-and-certification/docs-machine-learning-engineer-associate/AWS-Certified-Machine-Learning-Engineer-Associate_Exam-Guide.pdf.
As the exam is currently in beta, the material available for studying is limited. However the beta period is due to end soon and therefore it’s likely more study material will become available over time.
AWS Skill Builder has a bunch of content already available, with both free and paid for subscription options.
The free options are: -
- Standard Exam Prep Course (5.5 hours)
- Standard Exam Prep Plan (7.5 hours)
- 20 Practice Questions
- Partner Certification Readiness Course (29.25 hours, presumably only available to AWS Partners)
The paid for subscription based options are: -
- Enhanced Exam Prep Course (12.75 hours)
- Enhanced Exam Prep Plan (41.5 hours)
- Full 65 question practice exam
The AWS Skill Builder subscription is $29/month for pay as you go, so you could limit costs to $29 if you’re disciplined and have plenty of study time available.
There are a limited number of 3rd Party courses available at the time of writing, for example a number come up in a search on Udemy. Alternatively if you already have a subscription to a platform like A Cloud Guru you could wait for a course I assume they’re working on, or you could complement the free Skill Builder Learning with some overlapping content, for example the relevant sections of the Machine Learning Specialty certification course or some of their other AWS focussed Machine Learning courses that cover relevant exam topics or AWS services.
The exam uses the new question formats AWS introduced with AI Practitioner.
Ordering and Matching are both new formats, whereas case study is really a variation on existing themes, it’s simply that several questions asked related to the same scenario presented, rather than having separate scenarios per question. I have case studies with up to 5 questions in my exam.
As with all questions it’s really important to fully read the question and understand what it’s asking, for example where a question asks for lowest latency, highest availability or lowest cost. This will really help you when you get down to two of the available options that can be quite similar.
Based on my experience of the beta exam I’d say the following topics should be well studied, in combination with a good understanding of what the exam guide articulates. It’s especially important to be well versed in all the of Amazon SageMaker capabilities as I got a tested on those a lot.
- SageMaker — including experiments, pipelines, Lineage Tracking, Model Monitor, Model Cards, Autopilot, Data Wrangle, Clarify
- General knowledge of linear and logistic regression, including how to identify the right scoring system for a given use case (e.g. F1, RMSE, Precision, Recall etc.)
- General knowledge of ML theory — overfit, underfit, loss, model improvement techniques
- Generative AI (although not to the extent that SageMaker is tested) — knowledge of RAG, Agents, different models for text, images, impact of Temperature and Top-K
- Orchestration — when to use AWS Step Functions, SageMaker Pipelines, AWS Glue
- Storage services — S3, FSx (including Lustre and NetApp ONTAP), LakeFormation
- Data ingestion and preparation — Glue, Kineses, SageMaker Data Wrangler
- Networking — NACLs, SGs, VPC Endpoints
- IAM — roles, RBAC
In summary the AWS Certified Machine Learning Associate course and exam covers some interesting topics. If you’ve already got a good grounding of AWS AI ML services and general concepts some study to solidify knowledge on topics you rarely touch may be useful. If you haven’t got a background in AI / ML on AWS I would strongly recommend taking your time and doing a good amount of study to prepare for the exam.
