First, what is TensorFlow?

TensorFlow is an open-source library that is developed by Google and is widely used for machine learning applications. It provides a flexible and powerful platform for building and deploying machine learning models, and it is commonly used in both industry and research. One can build and train neural networks in just a few lines of code using TensorFlow. If this sounds interesting, you should keep reading.

The Certification

The TensorFlow certification is a way for individuals to demonstrate their proficiency in using the TensorFlow library to build elegant models to solve complex computer vision, natural language processing, and time series problems.

The certification is offered by the TensorFlow team and is highly recognized in the industry. Only a handful of developers over the world hold this certificate.

What's in it for me?

As I said, only a handful of developers can call themselves TensorFlow certified. So being a part of this exclusive group is really awesome. Apart from that, you get to showcase your skills on your resume and your LinkedIn profiles for the world to see. This can be a very attractive sight for anyone hiring TensorFlow/ML application developers. Also, all certificate holders get to be a part of the Certificate Network. Having such a huge network of ML enthusiasts can certainly help you in your career and the learning/collaboration opportunities are limitless.

The Exam

The TensorFlow developer certificate can be earned after passing a 5-hour online examination within a PyCharm environment.

The exam tests candidates on their ability to solve complex problems like Natural Language Processing, Computer Vision, Time series forecasting, etc. by building models using TensorFlow 2.x. It consists of 5 problems in increasing order of difficulty. The weightage of each problem depends on its difficulty. So, difficult problems carry more marks.

Can I use external resources like Google and Stack Overflow if I get stuck? Ans: Yes

You can use any resources you would normally use during your machine learning development.

What is the cost of the exam?

The exam costs $100 per attempt. Upon purchasing the exam, you have 6 months to appear for it before it expires. TensorFlow also supports a limited number of applicants every year by partially covering the cost of the examination. More info can be found here.

Note: The certificate is valid for 3 years from the date you receive your digit badge.

Great, how do I prepare for it?

You may have understood by now that this is unlike any other certification program where you watch 5 hours of recorded lectures and appear for an MCQ-based test. This certification is designed to truly test your problem-solving skills as well as your ML fundamentals. So, I would highly suggest you go through the following resources and prepare well before appearing for the test:

1. Machine Learning Specialization by DeepLearning.AI and Stanford University for Machine Learning fundamentals

   

2. Once you have your fundamentals right, you should start with the Deep Learning specialization by DeepLearning.AI [Highly Recommended]

   

3. Now, you're ready to think about TensorFlow and the developer certification. So, first, get yourself familiarized with the TensorFlow Certificate: Candidate Handbook [Highly Recommended]

   

   This handbook serves as the perfect guide for anyone willing to get certified. It contains all the information related to eligibility, registration, result declaration, fees, etc. I will highly recommend you go through this carefully before moving ahead.

4. The next task is to familiarize yourself with TensorFlow. TensorFlow Developer Professional Certificate [Highly Recommended]

   

   This course is MUST for every candidate. The course syllabus is completely in sync with the TensorFlow exam syllabus and I am pretty one of these was created keeping the other one in mind. 😂

   Also, if you're lucky, in your test, you may get an identical question that you solved in this course lectures or in the programming exercises. (At least I did 😎)

1. Intro to TensorFlow for Deep Learning course on Udacity

   

There are many other great resources like courses, books, etc. and you may want to take a look at them. However, I only stuck with the top 4. Whichever resource you follow make sure you study well and understand how it all works behind the scenes. This may be useful if you want to fine-tune your model sometime.

Ready for the test? Great. Let's see what's next!

Taking the exam

Once you're all set for the test, go to the Setting up guide to set up your examination environment. This contains the detailed steps for you to get your exam environment up and ready.

Note: Please install the latest version of PyCharm and not the 2021 version as mentioned in the setting-up guidelines. The 2021 version won't work. I wasted an entire day before I came to know this.

Once, you're all set, start the exam on your computer. Solve each question and submit your model for evaluation. Each model will be graded out of 5 points and you'll know within a few seconds of submitting if your model needs more work.

But what if I don't have a GPU? How will I train deep neural networks on my computer?

Don't worry. You can use Google Colab for training the models. Just download the .h5 file of that model after training and upload it in your exam environment.

Some tips and tricks

1. Write code snippets for frequently used callbacks, optimizers, and data augmentation. This will save you a lot of time and give you the freedom to think about other important parts of the problem.

2. You can also keep your programming assignments from the Coursera course handy and refer to them for syntax and formatting whenever needed.

3. If a model is scoring 3/5 or 4/5, fine-tuning a few parameters will most likely do the trick. Try increasing/decreasing the number of layers, neurons, epochs and batch size. Alternatively, you can also experiment with different Loss Functions like RMSProp, Adam, SGD, etc. Some loss functions will perform better than others for a specific task.

4. PRACTICE. PRACTICE. PRACTICE. The more practice you have, the more confident you'll be.

Just focus on learning new concepts and have fun while preparing. Follow all the things mentioned above and I'm sure you'll be well prepared for the exam in no time. Sharing my certificate below to serve as a motivation for you. Hope it helps. All the best!