GCP Machine Learning Engineer
June 8, 2024
Two years ago, I got the AWS Machine Learning Specialty Certification. It was a challenging exam for me to prepare for as a Program/Product Manager and I spent a couple months of solid prep time on it. I am using GCP more at work now so it seemed like a good time to pursue the GCP equivalent: the Google Cloud Platform Machine Learning Engineer certification.
The instructor-led training followed the same material as the Google Cloud Skills Boost Digital Cloud Leader path but was scheduled over 2 eight-hour sessions. I have some experience using GCP (mostly BigQuery) but have been using mostly AWS for the past three years, so I viewed this as a refresher. In retrospect, I would have preferred to take the Google Cloud Skills Boost training on my own time so that I could skip sections I was more familiar with. If you are new to Cloud technologies or GCP specifically, I would recommend covering all the material.
Since I have taken both exams and went into the GCP version influenced by my AWS experience, I will share with you the differences I found. AWS focused much more on general ML knowledge, while the GCP version was entirely on applying GCP solutions to given use cases. I spent more time than I needed brushing up on Metrics, Regularization types, and differences between Neural Networks! I will also say that I found the GCP exam much more challenging, although I did pass on the first try.
I spent about two months preparing for the GCP MLE exam and I think I could have done with a little more study time to feel less stressed during the exam itself. I am quite familiar with Vertex AI and BigQuery but still found a lot of questions challenging and saw questions on a wide range of topics, including Dataproc, Dataflow, containerized training, attribution techniques, and machine type selection.
I reviewed quite a few different materials to prepare for this test and will share with you both what I would and would not recommend.
What helped:
- Study Guide: Official Google Cloud Certified Professional Machine Learning Engineer Study Guide (Sybex Study Guide) by Mona Mona and Pratap Ramamurthy
- Google Cloud Skills Boost Course: Machine Learning Engineer Learning Path. . The MLE pathway is extremely long! I recommend completing all the labs but pick and choose the material based on your experience. I also found it helpful to use ChatGPT to summarize transcripts into study notes for me.
- Google Official Practice Exam: Cloud Digital Leader Sample Questions. Other practice exams I encountered or purchased were not as similar to the actual format as the official practice exam. Familiarize yourself with this format of questions!
- Google Cloud Machine Learning Engineer Certification Prep: Udemy course by Dan Sullivan. Good for a quick high level review in addition to the Cloud Skills Boost path.
- GCP Machine Learning Engineer Certification Preparation Guide: blog by dzlab
- Seven Day Study Plan: Cracking the GCP Certified Professional Machine Learning Engineer Exam in a Week by Zabir Al Nazi Nabil
- Multiple Resources: This helpful repo has rounded up many useful links
- A general tip as you answer the multiple choice questions: Always try to choose a solution in the order of GCP ML APIs whenever possible, followed by AutoML Training, Vertex AI inbuilt algorithms, Vertex AI Custom Training on TensorFlow, then Vertex AI Custom Training on containers as a last resort.
What was not helpful:
- Google Professional Machine Learning Engineer Tests GCP MLE: These Udemy practice examswere terrible. Many duplicates, poorly worded, and some were contradictory.
- The dozens of flashcards I made covering general ML topics (model evaluation metrics, activation functions, etc.)
- Some of the Google Cloud Skills Boost labs were frustratingly out of date (referencing AI Platform instead of Vertex AI, for example). At the time you read this, the labs may have been updated but in May of 2024, I found the Pipeline labs particularly out of date.
The exam itself was 50-60 questions long (mine was 50 questions) and you have 2 hours to complete the exam. It costs $200 and you can choose to take the exam in a testing center or via at home proctored exam. I took the exam at home, but if you are not familiar with at home proctoring, I would recommend going to a testing center for more ease and less stress. It took me about 90 minutes to complete the exam and review all the questions I had flagged (I marked around almost 20 for review!). After you pass your exam, make sure to check out your Google Certification Portal to claim your Google Cloud Certification Merchandise (I got the fleece jacket) and 50% off your next certification exam or renewal.
For someone like me that doesn’t spend a lot of time working hands-on as a Machine Learning Engineer, this exam was rough. If you work with Vertex AI on a daily basis, you will probably be fine with a few weeks of studying. This was a very satisfying exam to put behind me!
