About re:Invent 2018
Some people go to re:Invent to collect as many t-shirts, stickers, glowing swords, fidget spinners and socks as they can. Seriously, I've seen people walking around the Expo with multiple bags stuffed full of all the swag they can carry - these guys must come out to Vegas with an empty suitcase!
Not me though; I go to re:Invent to sit in quiet exam rooms, take beta exams, and then write blogs about the experience, making sure not to violate my NDA as I do so.
This year's new exam is the Machine Learning Specialty (or "speciality" as we might call it in the UK).
What is the Machine Learning Specialty Exam like?
I had seventy questions. I assume that the live exam will have fewer. Three hours in duration - again, I assume that the live exam will be 170 minutes in keeping with the other specialty exams.
Content-wise, I noticed an overlap with the Big Data exam in terms of ingest and storage of data, naturally enough - we're using large datasets to learn our models after all.
Also some questions around visualisation. I remember there being a few similar questions in the Big Data beta but I don't know if they made it into the live version. I passed the beta, so why would I sit the live exam??
The exam guide led me to believe... Let me rephrase that... After reading the exam guide, I got the impression that there would be a lot of questions on Amazon ML, not too surprising given the length of time that AWS take to produce a new exam, but there were in fact plenty of questions about SageMaker and even some around combining the two services.
Data preparation questions aplenty - strategies for splitting your data into training and validation sets and lots and lots and lots of questions on which algorithm, activation function, loss function, etc. to use for which use case. Also, not surprisingly, questions around infrastructure and security.
How should AWS users prepare for it?
Like all AWS exams, the exam guide is really important. That guide is the very first thing that AWS produce when they're building a new exam, so every single question can be tied back to it; it's the most important preparation aid of all. Of course Big Data on AWS and Deep Learning on AWS can also help you to prepare, but you'll need to know your SageMaker, Amazon ML and TensorFlow as well, none of which are covered in AWS instructor led training at the time of writing.
The exam guide says 2 to 5 years' experience with machine learning. I've been working in that area (obviously not exclusively - I wear a lot of different hats in this job) for around 18 months now and honestly, I'm not convinced that I'll have passed it when they finish their number-crunching in "around" 90 days. We'll have to wait and see...
Daniel joined QA in 2006, having previously worked first as a developer and then a trainer on the Microsoft stack. He is an Authorized Amazon Instructor Champion and holds all of the current AWS certifications. As a Learning Consultant, Daniel focuses on creating and delivering courses about cloud services, service-oriented architectures, software development, DevOps and data engineering.
Daniel also delivers our Google Cloud Platform courses, and holds 2 GCP certifications: Data Engineering Professional and Architect Professional. Other areas of expertise include: C#, .NET and agile development.
His areas of interest also include Microsoft Azure, Python, sailing, skiing and cycling – although not necessarily in that order or at the same time.
More articles by Daniel
5 good reasons to get cloud certified
Kubernetes certifications: QA's new cloud native courses
The benefits of AWS certification
AWS re:Invent 2016 - Certified Advanced Networking - Specialty
AWS re:Invent 2016 - Certified Big Data - Specialty
Converting CloudFormation JSON templates into YAML
A first look at the new Migrating to AWS course
The benefits of the cloud and Amazon Web Services (AWS)
Dude, where's my Data?
The Advanced Networking on AWS Exam, now Proctored by PSI