How did I clear the TensorFlow Developer Certification?
Preparing for this certification was the most rigorous technical studying, I have undertaken since my HSC board examinations. I have dedicatedly been studying for 1 hour every day for the past 3 months to complete both my Deep Learning Specialization and this certification.
To give you a bit of a background, I have previously worked on machine learning problems but I was somehow intimidated to try my hands at deep learning. This is when I learned about the ML Bootcamp organized by the Google DevRel Team and I thought it was the perfect opportunity to finally dive into the world of deep learning. The ML BootCamp sponsored the entire cost of the specialization and the certification exam for an entire cohort of students including me, so shoutout to the team for this amazing undertaking!
Unlike other certifications available wherein you could simply find answers to MCQs over the internet, the Tensorflow Developer Certificate requires a considerable amount of effort preparing, and like me, if you have not been working in the machine learning domain and want to be recognized by companies who design and solve real-world problems using deep learning, this certificate could actually be a good start.
Now that I have convinced you that the certificate cannot be passed merely by scanning over some material, let me share some resources that helped me get through the process of being certified by Tensorflow.
Understanding the basics of Deep Learning
To get a good hold on my deep learning basics I completed the Deep Learning Specialization offered by Andrew NG on Coursera. The specialization is comprehensive and covers the details of almost everything you would need for this certification and otherwise while working on a deep learning use case.
I am a visual learner and the below youtube channels have helped me a lot to concretize my understanding of some deep learning concepts.
https://www.youtube.com/c/deeplizard
Coding practice
As we have already established before, you need to code the model architecture in real-time during the examination which is why you would need an adequate amount of practice coding actual models. What greatly helped me is the TensorFlow developer professional certificate offered by DeepLearning on Coursera.
The course involves practical coding exercises in every sub-section and helps you to get hands-on with coding in TensorFlow really quickly.
Next, I participated in multiple Kaggle competitions to gauge how my models performed for a particular scenario. You would actually learn the nuances of deep learning like how to prevent overfitting, efficient use of callbacks, and how to use activation functions appropriately when you take the time out to sit and code.
When you learn a theoretical concept say a Convolution Neural Network, try to find an image classification dataset on Kaggle and code your solution for the problem, it would definitely increase your confidence. Combine theory with code and you are ready to go!
Below are some of my own public notebooks on Kaggle that I am sharing for inspiration. :)
- Using neural networks for a simple classification task:- https://www.kaggle.com/code/aliphya/competition-notebook-with-public-auc-0-99572
- Using transfer learning and CNN to classify images:- https://www.kaggle.com/code/aliphya/using-transfer-learning-to-classify-intel-images
Lastly, the certification requires coding in Pycharm, so for about two weeks prior to the examination I practiced coding deep learning models in the Pycharm environment, and believe me it helped me a lot during the actual exam. You certainly do not want to spend your time figuring out the exam environment during the exam, get acquainted early on.
That’s about it I guess, for this article. I hope I have earned the privilege of your time :)
