Google Cloud Professional Machine Learning Engineer Exam: Impression and Advice
On 01/15/2021, I took the Google Cloud Professional Machine Learning Engineer Exam. After submitting the exam, the website immediately showed that I PASSED đ€©. Google will need 7â10 days to review my exam record with exam terms and conditions. After completing their evaluation, Google will send me the final result by email.
(Updates: I got the official certification on 1/22/2021 đ±âđ)
While Iâm still waiting for the final email, I decided to write down my impression of the exam before I forget any details. I hope this can help anyone who is interested in getting the certification.
Exam Format and My Experience
You can take the exam either at a test center or from a remote location. I chose to take it at home because the nearest test center is still a 1+ hour drive from me. You can check the test center location HERE.
The exam is 2 hours. It has 60 multiple-choice questions. Because I was taking the exam remotely, I have to work with the test facilitator to check my identity and my test area prior to the exam. It took me about 40 minutes to complete the pre-check, but I believe my experience is not common. During the first check, they found my microphone did not work properly, so they asked me to fix it and rejoin the exam. Then, I waited another 20 minutes until the next facilitator was available to continue the check. The system will only start to count the exam time after you finish the check and read all the exam terms and conditions. Hence, you donât need to worry if the check will affect your exam time.
Personally, I feel the length of the exam is sufficient for answering 60 multiple choice questions. Under each question, thereâs an option box for reviewing the question later. This helped me a lot as I didnât need to spend too much time focusing on one hard question and worry that I donât have enough time for others. In the first round, I was able to answer most of the questions and marked some which I was not certain about. The first round took me about 60 minutes, and I marked 7â9 questions for review. Then, in the second round, I only focused on the marked questions. I read each marked question and all the choices word-by-word to make sure that I fully understand it. In total, I spend about one and a half hours completing the exam.
Exam Preparation
Now, a little background about myself. I have two years of experience as a machine learning engineer on GCP. Iâm familiar with the process for preparing, training, deploying, and monitoring an ML model on GCP, and I have experience using TensorFlow, Docker container, and CI/CD tools. Even so, Iâd say some of the topics on the exam was still new to me. Also, it was the holiday season which was not the best time for studying đ Hence, I spent about 20 days preparing for the exam overall.
Exam Preparation Guide
The exam guide provided by Google is useful for tracking your studying progress. If youâre looking for more detailed instruction, I personally found the preparation guide shared by Dmitri Lerko is very informative. Also, Google provides a Machine Learning Crash Course, and it is very useful for me to review the basic knowledge and strategy, like L1 and L2 regularization, on implementing ML.
Recommended Video Tutorials
I watched some videos provided by the Google Cloud (I use Pluralsight, but I believe you can find the same course on Coursera) :
- MLOps (Machine Learning Operations) Fundamentals: This course introduces participants to MLOps tools (AI Platform Pipeline, Cloud Build, Kubeflow, TensorFlow Extended, etc.) and best practices for deploying, evaluating, monitoring, and operating production ML systems on Google Cloud.
- Feature Engineering: Discuss good vs bad features and how you can preprocess and transform them using Dataflow + TensorFlow Transformation for optimal use in your models.
- Introduction to TensorFlow: You will learn how to design and build a TensorFlow 2.x input data pipeline. You will get hands-on practice loading CSV data, NumPy arrays, text data, and images using tf.Data.Dataset. You will also get hands-on practice creating numeric, categorical, bucketized, and hashed feature columns. This course will introduce you to the Keras Sequential API and the Keras Functional API to show you how to create deep learning models. It talks about activation functions, loss, and optimization.
Study Notes
Finally, I wrote down some notes while I was studying, and I shared them on Medium. I split the notes into sections following the exam guide. You can find my notes here:
- Section 1: ML Problem Framing
- Section 2: ML Solution Architecture
- Section 3: Data Preparation and Processing
- Section 4: ML Model Development
- Section 5: ML Pipeline Automation & Orchestration
- Section 6: ML Solution Monitoring, Optimization, and Maintenance
Taking notes helps me to double-check if I truly understand a concept by summarizing and paraphrasing it in my own language. Sometimes, I also just copy-paste the whole paragraph which I think itâs important. I was taking the exam in the afternoon. On the same morning, I skimmed all the notes for the final review, which made me feel more confident as well. Anyway, I hope my notes can provide you with more guidance and save a little amount of your time on finding studying materials. I know everyone has their own studying style, so I wonât encourage you to just follow my instructs. Instead, I hope my notes can give you some inspiration on scheduling your own plan. Also, please feel free to let me know if you find any typos or anything wrong in my notes.
Sample Questions
Google provides 14 sample questions to get you familiarized with the exam format. I took the sample questions after I completed my first round of studying. I made sure that I understood all the questions and each of the choices. For example, if the answer to a question is A, I will ask myself why B, C, and D do not work. I wrote down all the topics that I was not certain about. Then, I focused on those topics during my second round of studying.
Personally, I would say that the sample question is slightly harder than the actual exam. However, my feeling could be biased since I paid extra attention to study the topics that I was not familiar with after taking the sample questions.
Personal Advice
Preparation
- Make sure you understand the difference between each tool provided on GCP. For example, when do you want to use BigQuery ML vs ML API?
- If you have time to do the hands-on labs, do it! Youâll get hands-on practice in those Google Cloud video tutorials. It can assist you to get a deeper understanding and impression of the knowledge and tools.
- Make sure you are able to focus for two+ hours. This might be my own thing, but I always step away from the laptop and do some stretch every hour during my Work From Home time. However, you are not allowed to do so during the exam (of course đ). Hence, it was getting hard for me to concentrate after the first 1.5 hours (40 mins pre-check + 50 mins exam time). I even had a mild headache after I submitted the exam đ«. But again, this is just my own thing. All I suggest is that you probably want to train yourself to be able to focus for hours for the exam.
Taking the Exam
- If you are taking the exam remotely, making sure you read their instruction immediately after you registered. I didnât know that I was not supposed to use the monitor/keyboard/mouse until 10 minutes before the exam. It was a little rush to clean up right before the exam.
- Utilize the âreview the question laterâ function, but do not overuse it! The review option can help you manage your exam time more efficiently. However, if you mark half of the question, you might even lose more time on clicking back and forth.
This is all I want to share about my personal studying strategy and impression on the exam. My goal for taking this Google certificated Machine Learning Engineer exam is not about getting the certification. Instead, I use it as a motivation to encourage myself to keep studying new advanced tools and concepts of data science and MLOps. I hope you will enjoy studying and wish you the best of luck in your ML path!
