The test for the AWS Machine Learning Specialty is fairly difficult. Someone who has already passed the exam, such as me, can guide you. I’ve had to deal with some difficult questions. That’s why I decided to guide you so that you can easily prepare for it. The following are some suggestions for you to explore.
Understand the distributions: This is the foundation of statistical and machine learning knowledge. These will be put to the test, and you should be able to reply quickly. The question will give you a scenario to explain and will ask you to select the best distribution to describe it. I thought these questions were the easiest for a statistics student, but it doesn’t mean you shouldn’t obtain full scores on them because other exam takers are likely to get full marks as well.
AWS Kinesis Family: You may not have worked with real-time data before, which is why the AWS Kinesis set of tools exists. The kinesis family of services, along with AWS Glue, will account for the majority of the data engineering domain (20 percent) of the ML test. I’ll highlight a few big concerns.
- Only the Kinesis Data Firehose can load streaming data into S3, as well as compress and convert data to Parquet/ORC (for S3).
- Kinesis Analytics will employ IAM rights to access streaming sources and destinations.
- Another alternative is S3 analytics, which is not to be confused with Kinesis Analytics. S3 Analytics is used for storage class analysis.
AWS Glue: AWS Glue is another AWS service that I like to use because it searches my S3 parquet files and generates a schema for me. When we had a large amount of data in Parquet hosted on S3, querying Glue produced Athena tables for data exploration was quite useful. It features an AWS proprietary algorithm called FindMatches ML, which detects potential duplicated records even if they aren’t identical, in addition to standard data transformations like DropFields, filter, and join. Glue will construct elastic network interfaces, allowing the job to connect to other resources safely.
Security: Again, your mileage may vary, but I’m more accustomed to IT telling me I can’t do something because of security concerns rather than having to deal with the issue myself. You don’t want to lose any points on the ML test because there aren’t as many questions on security as there are on the solutions architect exam. At the very least, you’ll need to know how AWS S3 security works, as well as how to protect your data as it enters and exits SageMaker.
SageMaker: I’ve previously used Amazon SageMaker and found it to be useful for simplifying a few tasks, such as pre-installing the most prevalent machine learning frameworks in containers. I simply dislike the pricing, but that is beside the issue. Amazon SageMaker, an AWS fully managed machine learning service, will be heavily scrutinized in the ML exam.
Final Thoughts: That’s all there is to know about AWS Machine Learning Specialty Exam. If you want to learn more, there are thousands of AWS tutorial videos on YouTube to help you out. Use Career Ninja‘s Learn Tube chrome extension for hand-holding training on YouTube. Learn Tube organizes the results of your YouTube search into a course framework. If you type “AWS training,” search that term on youtube, LearnTube will show you a bunch of videos like an online course. As a beginner, you’ll click through the videos from the first to the last, as if you were taking an online course tailored specifically for you on YouTube.
