How MLOps and DataOps differ from regular DevOps

A question that we often encounter in any sort of technical industry in general is: what is the difference between a data role and a non-data role? What would be the difference between a software and data engineer, a data analyst and an accountant, or a DJ and a music composer? It is something employers ask a lot; people speculate on whether one is a subgroup of another or whether they are completely different. Even in the Swedish language, dator means “computer,” science is translated as vetenskap, and computer science is referred to as datavetenskap, so at some point whatever entity that designs and updates the Swedish language thought that there was very little to distinguish between the two.

We will now explain this through a couple of common DevOps use cases that can be applied and used in these more narrowed fields of DataOps and MLOps. For DataOps, we will go through a method that is simple but has saved me a lot of data concatenation operations in Python when using JSON files. For MLOps, we will focus on the GPU side, which is the primary hardware that an MLOps engineer may have to work with.

DataOps use case – JSON concatenation

This is quite a simple little trick, but one that is not as commonly known as you think. I honestly think if I can help even one person working with data with this section, I will have succeeded. The manipulation of JSON is a very important aspect of data operations that is very prominent, especially in NoSQL use cases, but also in a number of other cases. The ability to naturally manipulate JSON gives Python a major advantage over a lot of other programming languages. One of the most useful applications of this is the pipe (|) operator. This little operator can be used to perform concatenations, unions, and even bitwise operations on numbers. It is one of the many ways in which Python has made it easier to perform these small data operations for ease of use.

So, we will begin with just the function for the concatenation of one JSON with another:

a = {“one”:1, “two”:2}b = {“one”:”one”, “two”:2, “three”:3}print(a|b)

That’s it. That’s the code, and here’s the output of that code:

Figure 11.1 – Output of JSON concatenation

You’ll see that the second JSON’s value for a key overrides the value from the first JSON and that if they have the same common values, they will stay the same, and any additional values are combined into the overall JSON. So, with all that in mind, whenever you encounter such a problem with JSON combination (and it can come up quite often), you will have this little trick in your toolbelt. Now, let’s move on to another trick, one that will certainly help all you gaming hardware addicts out there. It’ll help the rest of you out too, but I like mentioning hardware addicts because they make the most YouTube videos and I’m hoping to get some of that sweet exposure.