DO's and DONT's while preparing for Amazon Machine Learning Specialty

Do you feel like you want to learn a new IT skill to unlock new opportunities and enhance your ability to innovate, rapidly prototype and develop some cool stuff? I have a suggestion for you - check AWS Certification Paths. In this post, I will share my story on how I got 3 AWS certifications in 6 weeks, including the new Machine Learning Specialty.

So why AWS and not something else?

 I recently joined Accenture and had what's called a "honeymoon" month. A time provided for new joiners to adapt to a new workplace, network with colleagues in the office, engage with business development and learn new skills. 

I've decided to invest my time and pursue the AWS Machine certificate that was on my to-do list. This winter I took the Cloud Technology course at UCLA Anderson, during which I used the Amazon SageMaker service to work on a Big Data group project. I was blown away by the processing speed and ease of development and prototyping: Need a server? Here, select one and it's ready for you in 2-5 minutes! At my previous job just a server delivery was taking 3-6 months :))) LOL

AWS is a market leader in cloud technology with market share significantly higher than Google and Microsoft Azure. To no surprise, AWS certifications are in strong demand and considered the highest paying in the US and Canada.

Amazon has several training paths for different role levels. I've decided to start with the foundational training before jumping to more advanced specialty certifications.

Source: AWS Certification 

Step 1. Certified Cloud Practitioner (CCP)

This certificate is targeted for Sales, Support, Managerial roles who need a basic understanding of AWS capabilities. It doesn't require previous experience with AWS. The course introduces different AWS software offerings, focusing on core services such as EC2 (compute), S3 (storage) and how to build secured, scalable, fault-tolerant applications on the cloud. You learn about cloud security and steps involved in migrating an on-premise solution to the cloud, how to estimate the cost for a new solution and differences in various support plans.

If you already have more than 1-year of experience on AWS and understand different AWS services, you might skip this one.

How to prepare:

1. Register on AWS.Training (free), take official course "AWS Cloud Practitioner Essentials" 4.5 hours
2. Register a Free Tier account on AWS to get some hands-on experience. Click Sign Up
3. Take an ACloud.Guru course "Certified Cloud Practioner"
4. That's it! Take a practice test and schedule an exam using AWS website

It took me 1 week to prepare. A Cloud Guru (ACG) is an outstanding platform and you can try it for free for the first 7 days. The test is very easy and I have high doubts that anyone would fail after completing all the steps outlined above. The cost of the exam is $100, and the duration is 90 minutes.

Step 2. Solutions Architect Associate (SAA)

 This certification tests your ability to integrate different AWS services into operationally effective, robust and secure solutions. AWS offers over 100 services and it is continuously growing by adding more and more capabilities and regions. I found this course interesting and engaging. It is like solving a puzzle and bringing all the pieces into one cohesive vision.

The course is not very deep. There will be no questions like "Select default retention of Kinesis Data Stream Service", but rather situations where you need to select the right type of service and implementation type. AWS offers the Solutions Architect Professional course that goes much deeper.

If you complete the CCP exam, you should already be familiar with different services. By taking this course you are going to take your knowledge to the next level and learn how to use different AWS products and services. For example, how to launch a web application and store customer data, create a VPC and secure your solution.

But don't worry if you haven't done this ever before. Below are the steps to prepare for the exam even if you don't have any previous AWS experience:

1. Take an ACloud.Guru course "Solutions Architect Associate". It includes a lot of hands-on labs (12.5 hours)
2. Read white papers recommended by Amazon
3. Read FAQ on S3, EC2, RDS, ELB, Route 53 and Amazon documentation on VPC
4. Take the simulation test on ACG. If you score 80% or above, then schedule the exam!
5. If you feel that you need more preparation then I suggest you look at the "Prepare for AWS Solutions Architect (Associate) Certification" path on LinkedIn Learning. I selectively watched some of the videos on security & encryption and it helped on the exam
6. If you want to practice more questions, check out Whizlab. I found their questions to be more similar to the actual exam's. 7 full tests for less than $20

 It took me less than 2 weeks to prepare, but I had some long nights. I heard that most people take typically a couple of months. Preparing for this exam was a lot of fun. I really enjoyed building Alexa skills. The cost of the exam is $150, and the duration is 130 minutes.

ACG has a lab where you provide test notes, then use Amazon Polly to convert text to speech and then build an Alexa skill to read your notes when you ask for it. How cool is this? Now I have some ideas on how to build fun answers for the never-ending questions coming from my kids!

Step 3. Machine Learning Specialty

 AWS ML Specialty is a very new exam and there is not much info on it yet. When I started preparing, ACG was releasing videos day by day and I had to check the course updates. It has since been completed and ready for you, but unlike the other 2 certifications, there are no courses that would teach you everything you need to know for the ML Specialty exam. Before you start preparing for this specialty you should be very familiar with the AWS console, how to launch new services, troubleshoot problems using CloudWatch, and configure secured access using IAM and VPC. If you completed Step 2, then you should be able to answer these questions.

For this course, you also need to have Machine Learning experience. Most of the questions on the test are not Amazon specific, but rather test your ability to implement an appropriate ML approach based on a business problem. I'll provide tips below on how to prepare for generic ML questions and get hands-on experience.

So let's start with what I suggest you to NOT DO in order for you to save time:

1. Don't read the documentation on Amazon Machine Learning Service
2. Don't watch suggested videos from Re:Invent 2017.
3. Don't watch all the official Amazon courses from Machine Learning path.
4. Don't take O'Reilly AWS ML course. I saw some people recommending it. In my opinion, it's a good overview of what to expect on the exam, but if you are going to take ACG course, skip O'Reilly, it's not going to teach you anything new

Now let's talk about what you should DO. I spent 4 weeks preparing for this exam, but a lot of time went into figuring out how to prepare and what to expect, so I hope this short guide will make your path much easier.

1.  If you are not familiar with different ML algorithms, I recommend starting with foundational ML courses. Luckily, there are many free resources nowadays. While Amazon suggests that 36% of questions are based on modeling, I found it to be closer to 50%, so knowing when to apply a particular algorithm and how to tune hyperparameters is extremely important. Below are a few courses on foundations of ML:

• EDX Machine Learning (not free) - I heard a lot of good things about this course, exercises are done using python
• Coursera Machine Learning (free) - The course focuses on the math behind the algorithms but uses Matlab Online for assignments
• Udacity Intro to Machine Learning (free) - The course includes the videos are less than 1 minute, so you can quickly narrow down to what you want to brush up

2.      Take the ACG course. It includes an overview of different Amazon AI & ML services and platforms that are part of the exam. The labs are outstanding! Big thanks to Cloud Gurus #KeepBeingAwesome

3.      Official AWS Training courses:

• "The Elements of Data Science" - It is very useful and several questions on the exam were based on the info from this video.
• "AWS Solutions Training for Partners: Machine Learning on AWS - Technical (Digital)"

4.      If you are not strong with Deep Learning check out Linked Learning course: Building Recommender Systems with Machine Learning and AI. There are many questions on the exam regarding neural networks.

5.      Read SageMaker documentation. You should know very well all built-in algorithms, when to apply each one and how to tune them. I created a cheat sheet for myself with a table summary of all built-in algorithms. It was very helpful on the exam because I could remember that spreadsheet and eliminate a lot of choices

6.      Take a practice test. It includes 20 questions and they are very similar to questions on the actual exam. If you score over 80% then go ahead and schedule your exam!

The cost of the exam is $300, and the duration is 170 minutes.

Exam tips

 Every time you pass an exam, Amazon provides a coupon for a free practice test and a 50% discount on the actual exam. You can actually save money by taking Solutions Architect and ML Specialty exams together, than just an ML Specialty exam ($300 vs $340)

If English is your second language, you can request an exam accommodation. Prior to scheduling exam, select "Request Exam Accommodations" and then from the drop-down choose "ESL +30 MINUTES". 

I really hope my notes are useful to you. Please feel free to ask any questions and let me know about your AWS adventure. Taking an AWS certification path is an exciting and challenging adventure. Good luck to all of you!