How I got Certified as a TensorFlow Developer by Google

Resources and tips that will help you prepare for the certificate exam

About a month ago, Deep Learning was a foreign concept to me — I barely had any theoretical background in it, and I had 0 practical experience coding neural networks. Now, a month later, I received the TensorFlow Developer Certificate, and I am pretty confident in building and training deep learning models using the state-of-the-art toolbox.

In this article, I am going to discuss my experience studying for the TensorFlow Developer Certificate exam. Since there are plenty of amazing resources and guides shared by other learners, and you probably have seen some already, I am not going to repeat them. Instead, I will focus on my learning strategy and exam-taking experience. I hope you find this article different but useful.

If you would like to read about other learner’s experiences, all the links are listed at the end of this article. They have helped me a lot, and I hope they can benefit you too!

1. Theory vs. Practice

Albert Einstein once said, “In theory, theory and practice are the same. In practice, they are not.” In machine learning, especially areas of deep learning, we find it hard to give definitive answers. Which optimizer should I use? What the learning rate should be? How many layers should I have in my neural network? Although we have come up with ‘rule of thumb’ to deal with certain scenarios, most of the problem solving involves ‘trial and error.’

But at the same time, only if we understand the theory behind a model can we properly apply suitable methods to construct and optimize it.

Do not rush to build a working model. Make sure you understand the mechanisms behind it and the correct ways to improve its performance. Please always keep this in mind while you are studying for this exam.

2. Reading and Coding

In my exam preparation, I forced myself to keep a healthy balance between reading theoretical concepts and coding.

In the morning, I usually watch several videos from TensorFlow in Practice Specialization on Coursera and go through some exercises that come with the course. In the afternoon, I read the Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow book, and try to mimic the code from the book (These 2 are the most helpful resources for the exam, there are more resources listed at the end of this article).

Everyone’s learning habit is different, so you should make your own plans. Just remember to balance between reading and practicing.

3. Jupyter Notebook and PyCharm

The exercises that come with the Coursera course are all in Jupyter Notebook and Google Colab. But the actual exam environment is in PyCharm.

For learning purposes, Jupyter Notebook is great since it allows you to see the output of one cell immediately after hitting ‘shift+enter.’ This feature makes it a lot easier for you to debug your code as well. However, in the exam, and in real life when you are deploying a machine learning model, you will be using some IDEs or text editors and code in an object-oriented way. Therefore, I strongly suggest you familiarize yourself with the PyCharm environment before the exam, practicing debugging in PyCharm rather than relying on Jupyter Notebook.