AWS Certified Machine Learning Specialty

 I must admit that the sense of accomplishment after clearing the "AWS Certified Machine Learning Specialty" exam and the adrenaline rush when you hit the submit button is slightly addictive!

This one is special because it is my first certification from AWS and being a DevOps Engineer, it would have been easier for me to take the "AWS Certified Solutions Architect" or the "DevOps Engineer" track instead of exploring less familiar terrain of "Machine Learning". However, stepping out of the comfort zone to learn something new made it even more fulfilling.

This post is about the learning path I followed in the run-up to this certification.

Machine Learning in itself is a vast field and this course, kind of scratches the surface and gets you started. 

In summary, you will need to know the following to clear this exam:

• How to identify the problem (Supervised, Unsupervised, Classification, Regression)

• How to choose the algorithm (Linear models, CNN, RNN, Tree Ensemble)

• How to train your model

• Data preparation & transformation

• How to use AWS ecosystem to solve the above

Distribution of Questions Asked:

The topic-wise weightage of the questions asked was as follows:

Topic	Weightage
Machine Learning Deep Learning	50%
AWS SageMaker	25%
About AWS Services	25%

• The total time to complete the exam was 3 hours

• There were 65 questions asked

Preparing for the exam:

1. I got started with watching `AWS Tech Talk` and `Deep Dive` videos on Youtube, not just about ML but about related services as well: https://www.youtube.com/channel/UCT-nPlVzJI-ccQXlxjSvJmw

2. Followed the free training videos and tutorials from AWS (not all of them though): https://aws.amazon.com/training/learning-paths/machine-learning/exam-preparation

3. ML/DL needs some high school/college level mathematics to be revisited. Basically, Linear Algebra, Probability & Statistics, Multivariable Calculus and Optimization, worked for me.

4. Data Visualisation using Jupyter notebooks.

5. Regression and gradient descent.

6. DL Models - CNN, RNN

7. Worked on understanding the following concepts-

1. Supervised, unsupervised and reinforcement learning.

2. Purpose of training, validation and testing data.

3. Various ML Algorithms & Model Types-

1. Logistical Regression

2. Linear Regression

3. Support Vector Machines

4. Decision Trees / Random Forests

5. K-means Clustering

6. K-Nearest Neighbours

Once the above concepts are understood go ahead with trying out the following AWS services-

• SageMaker

• Rekognition

• Polly

• Transcribe

• Lex

• Translate

• Comprehend

• S3 including how to secure your data

• Athena including performance

• Kinesis Firehose and Analytics 

• Elastic Map Reduce (EMR)

• AWS Glue

• QuickSight 

Those of you who regularly use AWS services won't have much of a problem grasping these.

Finally, try practicing a lot of practice exam questions like ones from the link below:

https://www.udemy.com/course/aws-machine-learning-practice-exam/

You should also have a go at the official practice exam before going for the mains. So that was it folks. I am still learning this discipline, and it's all volatile right now. I will feel more confident with ML once I start applying it in some real-world applications. Will write about those experiences as they come by.