How I passed the Tensorflow Developer Certificate Exam (and how you can too!).

Is it worth it?

This question is hard to answer. This depends on many things including your profile and what you’re looking for. Here are two reasons why it might be for you:

  • You want to acquire the foundational skills required to build ML-powered applications.
  • You want to stand out to a prospective future employer

Of course, a paid certificate doesn’t guarantee anything. It’s just another way to showcase your skills. You can acquire all the skills required for passing the exam without being certified.

5 hours, 5 problems

Read the TensorFlow Developer Handbook and you’ll get a fair idea of the major sections of the exam. Practising all of the concepts mentioned in the handbook will be very helpful. There will be 5 problems and each of the models will be marked out of 5. I’m not sure about the passing criteria, but I would say a total score of 20/25 is required. It is also mentioned that if you skip a question it will impact your chances of passing.

Preparation

Basics

Before you start building your models, you want to make sure you have your basics down. I would recommend familiarizing yourself with these before attempting the exam:

  1. Python fundamentals and some popular libraries such as pandas, numpy, matplotlib and scikit-learn.
  2. Reading the tensorflow documentation. This was my bible during my preparation period and I probably spent more time on the tensorflow website than any of the resources listed.

Courses

There are plenty of great, free courses available online to guide you through your preparation. The course that I referred to for my preparation is Udemy: TensorFlow Developer Certificate in 2022: Zero to Mastery. It was very helpful and comprehensive and covered all the things I needed to pass the exam. The instructor, Daniel Bourke has also passed the exam and written a blog on the experience.

Additionally, another popular course is Coursera: DeepLearning.AI TensorFlow Developer Professional Certificate. It should also be noted that this course is an officially listed course on the TensorFlow website.

Additional Sources

  1. Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow 2nd Edition
  2. MIT Deep Learning 6.S191
  3. Blog posts!

Tips and Tricks

  • Spend time looking for tutorials on the official Tensorflow core website. There are very thorough and contain a lot of relevant information.
  • Familiarize yourself with the IDE at least a few days before attempting the exam. PyCharm can be a little intimidating if you haven’t used it before.
  • During the exam, we can train the models using Google Colab. This allows us to take advantage of the free GPU provided in a Colab environment, which speeds up the training time. You can simply upload the .h5 file inside the appropriate folder for evaluation
  • Set up a test environment in PyCharm using the exact version of Python and packages that the guidance details prior to the exam. This should mitigate any potential problems you might face and avoid last-minute panic.

My experience

I had planned to take the exam on a Sunday(12/4/2022) so I bought the exam the day before. I double-checked that all my pycharm configurations were correct and consolidated all my study materials. I had devoted the past two weeks to exam preparation, so I was feeling pretty confident. From the information available on the website as well as other blog posts(like this one), I had a general idea of what to expect.

Fast forward to D-day. I decided that I would start the exam at 8 A.M. However when I pressed the “start exam” button, I was redirected to a page that said “setting up exam environment”. Underneath the progress bar, it said that the page would automatically refresh once it was done. The progress bar was stuck at 99% for almost an hour before I got impatient and refreshed the screen. Lo and behold, I moved on to the next phase.

I can’t share the details of the exam as that would be unethical but it took me close to 4 hours to finish the exam. This might vary with your experience levels as I’ve read other blog posts where people were able to finish within 2 hours. The first two models were a breeze, the third had a small catch, and the last two were tricky.

Once I felt confident with my performance, I went ahead and clicked the “end exam” button. Immediately I got two emails. One with the subject “Congratulations, you have passed the Tensorflow Developer Certificate Exam!”. Yes, I did it! A wave of accomplishment(and relief) washed over me. The second email had my badge and certificate, along with options to share them on different social media platforms. Within a couple of weeks, I was added to the Tensorflow Certificate network.