Sharing my journey from preparation to certification
After I finished my requirements for university, I reassessed myself and set up on a 4-month journey to study AI, following a curated bottom-up curriculum using the books that I bought and courses. Looking back, this is one of the best decisions I have ever made.
Shoutout to the slack communities that helped me to be accountable in this study — Udacity folks from AI Network and Filipinos from the former AI Study Group PH, and twitter folks.
One of the checkpoints I have included in my curriculum is to pass the TensorFlow Developer Certification Exam. I decided to take it as it challenges me since it is a hands-on exam, and it pushes me to actively study TensorFlow.
First, we have to set our expectations right. This is a different exam because it is hands-on type which can be taken online and needed to be finished in a 5-hour time limit. This sentence is too compressed, let’s elaborate.
Hands-on Type and Online Exam
You will be asked to have your environment ready and don’t fret, there’s a detailed instruction for setting up the environment for the exam . Follow it very closely, double check, especially the versions of the dependencies needed to be installed then breathe. I actually can’t count how many times I have done that.
The major platforms I used were Google Colab and PyCharm 2020.2. The use of Google’s Colab is not mandatory but recommended to take advantage of the GPU. This information is also in the guide. This also became a need for me as my laptop hangs and take some time in training neural networks. On the other hand, the specific version to be used for the PyCharm IDE is very important. Make sure it is 2020.2 as of this writing. By following the instructions closely, I did not experience any problems or errors while taking the exam.
5 hours for 5 problems, No Extensions
The problems along with the other information, which is enough to let you know what you have to do, will be given in the Exam Environment to be installed in PyCharm. There are two cases for which the exam can end. First, if you still have time left, you can choose to click the End Exam button. Lastly, if you have no time left, the exam will automatically end and all your progress will be considered. In my case, I ended it with some minutes left because I took more than 2 hours on that one number (the last number) because I want to get 5/5 and not 4/5. With regards to the first four problems, I finished it within 1.5 hours feeling satisfied with 5/5. So yes, you will know how good your model so far.
Exam Payment and Registration
The exam costs $100 and be prepared to ready a valid ID. You can pay for it and take the exam at a later time. Though, beware of the expiration of the access upon payment.You can find the process by clicking the Begin exam button found in the website.
Don’t worry, clicking that button won’t start the clock ticking for the exam because you actually haven’t paid yet. It will redirect you to this:
You can also avail the TensorFlow Education Stipend. This stipend offers various assistance you can take advantage of. The first one is the exam stipend which will give you 50% discount ($50 USD). The other choice is to avail both the stipend for the deeplearning.ai’s TensorFlow Developer Certificate and the exam stipend. You will be able to access all included courses via Coursera. Though, the exam stipend still offers only the 50% discount like the first option.
Take note that you should apply for this stipend early because it will take 4–6 weeks to get the results of your stipend application. Yes, results, because it will depend on your application if you will be given the stipend or not. In the application form, you will answer various questions where they will base their decision. If you made it, you must also take note of the expiration of the access. See more details of this offering here.
The best way to know the coverage of the exam is through the dedicated website for it. There are many things in there but I will highlight some that I used.
This sits on top of it all. I can’t emphasize that further. I have read this many times while starting, preparing and even right before the exam. This is like the battle plan being handed to you because it contains the Skills Checklist.
This candidate handbook, along with other materials, is being updated from time to time so make sure you are using the latest version which you can find in their website. I said that because you may find other blogs that actually used the old versions. So as of this writing, this was updated last October 26, 2020. I have read other blogs and so far, the new section that I noticed was the TensorFlow developer skills which is the first section.
deeplearning.ai’s TensorFlow Developer Professional Certificate
If the candidate handbook is your battle plan, this is your battle plan details. Trust me, if you have completed this, have coded this, have religiously studied this or really know this, you will certainly pass the exam.
I second this note written in Coursera:
This program can help you prepare for the Google TensorFlow Certificate exam and bring you one step closer to achieving the Google TensorFlow Certificate.
I keep repeating this because I want to stress that really. If you don’t have much time and you know the concepts of Deep Learning for simple regression, computer vision, natural language processing, and time series and just need to learn the TensorFlow part, take this.
- Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow 2nd Edition
This book is awesome. If you want to learn from classical machine learning to neural networks using one book, spot on, pick this book.
I bought this in Amazon during their sale last year. I was actually using this book for classical machine learning algorithms but for the exam, I only used it as a reference book. I can say, what’s written in the second part of the book covers a lot more than the exam. So if you want to focus on the exam first, you may just want to use this as a supplementary material. Surely, after the exam, if you want to know more, go back to this as your primary material. Definitely worth it. That simply explains the 5 star rating in Amazon.
Given my previous months of learning Deep Learning concepts, I used this only as a review and only picked the relevant stuff.
Reading Blogs (like you!)
Been there, done that. Learning and listening to people who have done the exam is such a great thing to do. These are some blog articles that I found really helpful (in no specific order):
- How I passed the TensorFlow Developer Certification Exam by Daniel Bourke
2. How to Pass the TensorFlow Developer Certificate Exam by Harshit Tyagi
3. Part II: The TensorFlow Developer Certificate by Viren Radhakrishnan
Apart from this, I also tried to look at different forums and reddit posts. Of course, this should be done ethically, in a way that it just helps you in the preparation but not cheating! In the first place, you’re doing this not to just get the certificate but to challenge yourself and learn.
Practice, Practice, Practice (Repeat)
This is one thing that will increase much of your chances more than anything else. I recommend practicing the exercises and the programming homeworks from the TensorFlow Developer Professional Certificate using Google Colab and PyCharm. I did that, I also practiced once more after each course and I practiced all of it again after finishing all of the courses.
Before the exam, I set aside one week to practice again and increase my familiarization with the workflows for the different parts like in Computer Vision, Natural Language Processing, and Time Series. Different data and use cases may arise but it’s one step forward if you understand and know the given use cases in all of those courses. I even made flowcharts for each use case and given different data and see if I can recall it well.
It’s not necessary to do it the way I did it. Just have a way to make it stick enough. When I was doing the exam, I thanked myself for preparing that way. It worked!
After you press that End Exam button, you will get a confirmation whether you pass it or not. This image below shows the snapshot of the email I received.
After less than 2 weeks, I received my certificate and I saw my profile being added to the TensorFlow Certificate Network. This is an interesting place where people looking for TensorFlow Developers can find you.
Passing the exam before New Year’s Eve added up to the excitement and the celebration going around. It’s a good year ender and the day after that, I started a new chapter for 2021. As of this writing, I am now working on AI and Computer Vision projects as a hired individual and also as a volunteer to other impact projects. Learning just never stops.