Academic Data Science curricula are unsurprising but disappointing in their stubborn omission of cloud technologies. It’s expected that academia would stick to theoretical knowledge and avoid teach any vendor-specific skills, but it’s disappointing because basically every entry-level software engineering job requires these two things: experience working in a CI/CD framework and experience deploying production-ready code in the cloud, and neither of these two skills can be gained through school. If you’re lucky, you can get an internship which will give you this experience, but sometimes even internships can be hard to get and require prior experience… There’s this chicken-and-egg problem when starting your career in software engineering, and at the end of the day, most practitioners in the field admit that you just have to get lucky to get your career off the ground.
I’m not one to rely on luck. So even though I do have internship experience in Google Cloud, I’m taking the time this semester to work through one of Google Cloud’s online learning path: Data Scientist / Machine Learning Engineering learning path. This is one of many online curricula that Google Cloud offers including Data Analyst, Data Engineer, and Database Engineering.
Each of these learning paths are about the length of a traditional semester’s course work and contain a combination of Coursera online video lectures and quizzes with hands-on QwikLabs Skill Badge exams. Each is self-paced and can be done in my own time as if it was a semester-length course.
I’ve never found that online credentials have much meaning to employers, but I am enjoying filling in the gaps in my academic machine learning education.
So far, I’ve found that they complement each other rather nicely. (“they”, being both my Master’s degree and this Google Cloud’s learning path). Both are, on their own, sometimes a bit disappointing to me. Academic courses go into great detail about the history of machine learning algorithms and how they’ve been optimized over time, but they say almost nothing about what applications these models might be used for or how to deploy them to solve real-world problems. It’s as if they are afraid that the pure theoretical work will be tainted by the practical applications of these models.
And on the other hand, this Google Cloud learning path covers basically no details of the models. I can tell that I wouldn’t be gaining as much as I am from these courses had I not already learned the technical aspects of machine learning through the my courses at school. (This reminds me of Tyler Cowen’s famous answer to the question, “how long did it take to read that book”, to which he replies “it took me 45 years”, insinuating that the prior knowledge that you bring into reading a book is more important than how fast you can skim.)
Needless to say, I’m enjoying learning about the operations and the business context in which machine learning models operate. I’m a third of the way through the learning path (finished two Coursera courses and one Skills Badge), and I’m already thinking that after this, a deeper dive into MLOps and DataOps will be well warranted.
Stephen Gass 