Is the AWS Machine Learning Certification worth it?
How does AWS Machine Learning Certification Compare to the Competition?
Machine Learning is hot, and AWS machine learning is at the very top of the list. When compared to the other major cloud providers on SpyFu (a search keyword research tool), the term "AWS Machine Learning" leads in monthly search volume* when compared to the other four major cloud providers:
- AWS machine learning: 5,100
- Google machine learning: 2,100
- Microsoft azure machine learning: 1,200
- IBM machine learning: 840
- Alibaba machine learning: 125
*My search was conducted January 3, 2022
GoogleTrends, show different results, with Google leading the pack and AWS Machine Learning trending upward, however we could certainly raise some concerns about potential bias:
Gartners' Magic Quadrant for Machine Learning
According to Gartner, AWS (Amazon Web Services) ranks high in the machine learning "Visionaries" box, though they do (according to Gartner) trail Google and Microsoft:
In our experience as machine learning consultants it is more nuanced. Each of the major ML could providers have different strengths. For example, Google has a fantastic speech to text offering, while AWS' visual machine learning is top tier.
Should I get multiple machine learning certifications?
With different strengths from different cloud providers, one might be tempted to create solutions that span multiple ML providers, and/or to pursue multiple Machine Learning certifications. I would strongly recommend against this. The reason for this threefold:
- The amount of change that is coming in the ML space is massive - sticking with a single ecosystem makes it more manageable.
- All the major providers have pretty good offerings in most spaces.
- All of the services are constantly being improved.
When deciding which machine learning certification to get, pick the most important tool for your needs and focus on that. If in doubt, I personally recommend the AWS Machine Learning certification, as their overall certification program is the most robust.
What are the top AWS machine learning tools?
Within the AWS Machine Learning ecosystem, there are a number of products and services and there are new ones being released every year (here are some of the latest announced at AWS re:Invent 2021). The top 5 AWS Machine Learning tools (according to spyfu) are as follows, with their monthly search volume:
- AWS SageMaker 4,200
- Amazon Lex 3,000
- Amazon Rekognition 2,300
- Amazon Transcribe 1,000
- Amazon Comprehend 750
AWS SageMaker is Amazon's core Machine Learning product, which all other ML products are based on. It is Amazon's most searched for machine learning tool.
Technically considered an AI (Artificial Intelligence) product, rather than an AWS machine learning service, Amazon Lex is a platform for building chatbots.
Amazon Rekognition is a machine learning image and video analysis engine. With out-of-the-box capabilities for content moderation, face compare and search, face detection, sentiment analysis, custom labeling, text detection, and more, Rekognition is one of Amazon's most advanced machine learning tools.
Amazon Transcribe is AWS' speech to text artificial intelligence tool. Current specialized use cases include call center analytics, medical transcription, and real-time subtitles. While currently not as accurate as Google's speech-to-text, it is improving rapidly due to Alexa's broad distribution.
Amazon Comprehend is an artificial intelligence tool which helps extract insights from written text. Specific use cases include call center analytics, sentiment analysis, redaction of Personally Identifiable Information (PII), and more.
How much do Machine Learning Engineers make?
According to Indeed, Machine Learning Engineers in the United States make an average base salary of $131,000 per year. In my experience this would be for an engineer with a graduate degree and a number of years of experience in a related field (software engineering, etc.). Starting salary for someone with zero years of experience and no higher level education might be in the $60k to $80k range, but compensation could easily reach the six-figure mark with a few years of experience. It's just a matter of getting experience, and machine learning certification are great qualifiers.
Is the AWS Machine Learning Certification worth it?
While it can take at least two weeks or potentially several months of study to achieve (depending on your experience level), the AWS Machine Learning Certification is well worth the time investment. The ecosystem is robust and growing and the traditional education pathways are struggling to catch up. In the United States, only Carnegie Mellon University offers a bachelor-level machine learning program, and the need in the industry is massive. All other programs are masters or PHD level, which also means huge amounts of student debt.
By comparison there is a one-time fee of $300 for the AWS Certified Machine Learning - Specialty exam, and the practice exam fee is US$40. The test takes 180 minutes to complete, and preparation is generally in the range of 40+ hours. The official AWS partner machine learning training learning plan is supposed to take 34 hours and 40 minutes (something I'm currently working through myself):
How to prepare for AWS machine learning certification?
If you don't have access to the partner training provided by AWS there are a number of online platforms that provide preparatory courses. I have no affiliation with any of them and I haven't gone through them so I can't speak to their effectiveness, but my expectation is that they are all pretty good:
- Udemy: AWS Certified Machine Learning Specialty - 10.5 hours - $99.99 (on sale for $15.99)
- A Cloud Guru: AWS Certified Machine Learning - Specialty - 13 hours - $35/month
- Cloud Academy: AWS Machine Learning – Specialty Certification Preparation - 53 hours - $39/month
I'll be going through the official AWS training (35 hours worth) and then taking the exam without any other AWS certifications to my name. I am, however, coming to the table with:
- 22+ years working in the tech space
- a few university-level computer sciences classes under my belt
- 10+ years as a working web developer
- A few years of exposure to general machine learning concepts
Is a 35 hour crash course enough additional exposure to get the cert? I'll let you know how it goes!