Machine Learning
ML Engineer vs Data Scientist: Whom Should You Hire?
Not sure if a specific business need of your company requires a Data Scientist or an ML Engineer? Then read ahead to find out more about it.
Some time ago, a client approached me with a project proposal. He wanted me to use a certain Product-Market framework & write an article from a Data Scientist’s POV. His reasoning, “my Machine Learning skills could prove useful to analyze the drawbacks and/or gaps in the framework to add to it”. And to be clear here, I’m no Data Scientist by craft either.
A couple of emails later, it didn’t take me long to realize his requirements. His target audience for the article were individuals in a non-engineering field. Perhaps, marketing (my assumption) or maybe analysts.
It made me wonder, are my credentials on LinkedIn, GitHub or heck, even Twitter not visible to anyone? The fact that the prospect reached out to me with a data science task instead a software engineering job, points out to something. The issue of ambiguity in the roles of a Data Science/Machine Learning practitioner.
The responsibilities of a Data Scientist and an ML Engineer are often interrelated. Depending on the size of the company, it wouldn’t be surprising to see a Data Scientist working as “the Engineer”. In other words, startups, especially those in the early-stages, resort to hiring Data Scientists as the Swiss knife for all their analytics, data science modelling, productionizing ML models & other needs. Those were some common roles to name a few.
It’s common for a Data Scientist to be expected to wear many hats in the industry & that’s the root problem.
Although, as of writing this article, the trend within the industry is changing for the better. An increasing number of companies have realised the benefits of the Division of Labour. They’ve been hiring individuals with specific skill sets in either ML or Data Science. Not both.
Even then, some entrepreneurs are still not sure if a Data Scientist will be the right for an analytical job. The client who reached out to me was one such entrepreneur. But he wasn’t the first & willn’t be the last either. It’s because of individuals like them articles like this — Data Analyst vs Business Analyst. Here’s the Difference by Matt Przybyla exist. This article is proof of the need to define the various subfields of Data Science.
Hence, for the sake of the Machine Learning community & as David Robinson says in his tweet,
When you’ve written the same code 3 times, write a function
When you’ve given the same in-person advice 3 times, write a blog post
— David Robinson
Hence, this article on the topic of Data <something> role vs Machine Learning Engineer is my contribution to the community. By the end of the article, you will have enough idea of the respective responsibilities of an ML Engineer or the Data Scientist.
Why Does a Clear Distinction Between These Roles Even Matter?
The plain & simple answer to the question is Division of Labour. As Adam Smith among other prominent political economists of that time, before he & after as well, stated, the improved efficiency & increase in greater skill gained from division of labour is hard to pass on. From an entrepreneur’s perspective, this sounds like a free lunch.
Often businesses operate under tight financial constraints. Hence, having an ML Engineer focus on only one task, say ensuring the data science models are production-ready, saves both time & money. This way the engineer can take care of any blocking issues. And I’ll not have to worry about being that boss who switch contexts at a whim.
Beside the not-so obvious benefits from a business perspective, there’re other benefits as well. Some which you might or mightn’t have realized already.
A clearly defined set of responsibilities, work in favor of the employees to better understand the goals of the project. They can then set aside time & resources as required to achieve that goal. Besides, a major factor to consider is, not to burden the employees with overwork.
A defined designation of roles does have many obvious added advantages then. So if that’s how it is, what specific work should the Data Scientist and/or the ML Engineer be assigned to? In other words, how do you define responsibilities for that job position?
The next section attempts to answer these questions.
Who’re ML Engineers Or Data Scientists & What Do They Do?
This article is one of many similar articles on the same topic “Machine Learning Engineer vs Data Scientists”. Those articles often describe their respective responsibilities of those roles. But the descriptions are factual & only explain job descriptions, that’s it, nothing more & nothing less. None of those contents explain the specifics that define the two job responsibilities.
One such article I stumbled upon & also the one which inspired me to write this article is — Machine Learning Engineer vs. Data Scientist on the Springboard Blog by Andrew Zola. The article is factual as expected, but sheds enough light onto the differences. In other words, a perfect article, brief enough for anyone to quickly grasp the differences between a Data Scientist & a Machine Learning Engineer.
Just so you don’t have to go through the article, here’s a brief summary of the article.
The author mentions ML Engineers as individuals who leverage big data frameworks. Besides, their in-depth coding skills are also used to build intelligent programs. They’re also responsible for productionising theoretical data science models into practical applications.
Fair enough. That’s very atypical of any engineer from any background regardless of their domain.
To reiterate on his definition, the ML Engineer is like any other traditional software developer. But the only extra skills are those which are necessary to work within an ML environment. While Data Scientists are individuals who look into all aspects of the business, to develop programs (not necessarily software) as per the requirements.
In other words, a Data Scientist handles tasks more akin to what a typical analyst has been doing for ages. “Data Scientists” is just a fancy term (or designation) given to them only after the recent AI hype.
It’s not uncommon to find Data Scientists with only limited programming knowledge. After all, querying a SQL Database , making some visualizations using Python/R are often all the tasks they are ever expected to do. At most, they’re responsible for designing the data science model for implementation into a production-ready app.
At a glance, the two roles have very distinct & defined roles, yet employers don’t seem to care about them. Could it be a simple coincidence? Or maybe the employer didn’t know what skill set is needed to solve a business problem?
To get a better understanding of the situation, let’s take a look at some job descriptions. And the following section will do exactly that.
What Skills Are Employers Looking For & Their Job Descriptions
On paper, anyone can point out distinct responsibilities for specific roles. But there’s no saying it’ll be the same out there. Often job postings are put up by individuals who’re not aware about the intricate details of the roles. Or the entrepreneur could’ve delegated the responsibility of hiring to a recruiter. The recruiter then unknowingly lists out all those requirements under one job posting.
Weird, I know, but happens more often than you would realize. Don’t believe me? Then check out the screenshots below. They’re examples of actual job postings I picked up from Indeed.com while writing the article.
With that said, let’s take a look at some of the job descriptions for a ML project. Below is a screenshot I picked up from one such job ad posting.
At a quick glance, there’s nothing of significance in the screenshot above. Pretty normal for a job description for any traditional software development roles.
But on the flip side, job descriptions of Data Scientists are often vague & kinda similar to other software-related positions.
Check it out for yourself.
This screenshot is a prime example where it’s not clear what the Data Scientist is expected to do. It’s a hodgepodge of all sorts of responsibilities dumped on one single individual to take care of.
How Should You Adhere To Specific Designations?
There’s some opinionated advice in this section. And by no means should you feel obligated to adhere to this advice. But I expect my readers to make an informed judgement by themselves based on their needs.
- Designations don’t matter. Yes, you read it right. Reading the article till here, that statement must’ve caught you by surprise? But here’s why. For employees to wear many hats in a startup is often beneficial in the short-run. But as the manager or even the founder, expecting your employees to tackle many responsibilities is a bad sign. It signifies an inability to manage scarce resources. Most important of all it shows the founder’s inability to use the employer’s time properly.
- Be clear on the job description about the responsibilities the employee should tackle.
- Understand specific problem areas in the business & hire accordingly. Prospective employees with particular skill sets should be assigned to specific problems.
- It’s easy to switch contexts in a Data Science/Machine Learning business environment. The ambiguous nature of the fields puts the employee in a situation where their skills can prove useful in many ways. Avoid it at all costs, if possible. This is easier to remember when ML is seen as a sub-field of Data Science & not the other way around.
My final words.
Designations are mere names we give each other for the sake of simplicity. And they’re helpful in understanding the respective roles of operation of a business. What names are assigned to the roles don’t matter. What matters is, the work to be completed & shipped. Communicating those goals should be priority over specific designations.
And as entrepreneurs, be sure about the needs & requirements of the business. Communicate them to your employees properly.
