What’s in a title? A Data Science conundrum

Having recruited in the Data Science field for over 7 years, there is a something that’s been bugging me from the start of my career which continues to bug me today.

It’s a problem that I know is shared by many others in the industry, including scientists, engineers, recruiters and even entire businesses (and not just those that are hiring Data Science specialists - I’ll explain how later).

The problem is the fact that job titles in data science are unclear.

In this industry, one company’s Data Scientist can be another’s ML Engineer, Data Analyst or sometimes even wilder titles like Ninja, Unicorn or Wizard!

It’s kind of ironic that a sector that is full of logical thinking people, and which routinely makes sense of unstructured data, has yet to agree upon and implement standardized job titles.

Whilst Data Science companies are busy finding solutions for almost every other walk of life, they seem to have left behind the problem of simplifying the identification of their own roles. It brings to mind the sayings ‘a builder’s house is never finished’ and ‘busman’s holiday’. 

There are some RecTech startups that are going some way towards simplifying things, but as businesses selling their service to help identify people, it’s counterintuitive for them to provide a service that could essentially reduce their need in market.  

Why is this a problem? Well, for almost all other professions and trades, it’s very simple to identify what someone does. For example, if I were to say I needed a painter instead of a painter/decorator, vehicle sprayer or an artist, you’d be forgiven if you misunderstood what I needed them for, right?

However, in this market, it’s acceptable for a hiring manager to say I need a “mid-level data scientist with Python, SQL and R” or for a candidate to say ‘I’m interested in the role you posted: MLOps Engineer. Based on my experience as Senior Machine Learning Engineer, I believe I could be a good fit.’ (This was just copied from a message I received an hour ago)

How is an MLOps Engineer the same as a Machine Learning Engineer? And if it is, why are they not called the same thing? It’s akin to someone telling BA they’d be great at flying a passenger jet across the Atlantic based on their experience as a drone pilot. 

Absurd, right? Or is it? It’s hard to tell!

This is a problem for: 

Anyone looking for a new position in the industry

Without doing quite a bit of time-consuming research you probably won’t know company X’s Data Scientist position isn’t aligned with your experience. Unlike the companies you’ve previously worked for, their Data Scientists are required to do the work of Machine Learning Engineers. This leads to mismatched applications and rejections (or sometimes no response at all). The one-click application process exacerbates this even further as applicants often just read the title and click apply. 

In addition, it’s very common nowadays for people to be invited to interview only to discover that they aren’t anything like the right match for the role that they have applied for. Bearing in mind that any interview for a Data science role potentially involves a whole load of research, time off work, arrangements over care provision, potential travel and occasional stress, anything that helps to negate this scenario should be welcomed! 

Talent teams and recruiters

Whilst scratching your head wondering why a guy that’s laid data cables for Sky Broadband for the last 15 years has applied for your Data Engineer position (true story btw), the lack of uniformed titles makes it difficult to decide who to approach for your Machine Learning Engineer role, especially given that technical tools and frameworks often cross over between a lot of these positions. 

How do you decide who not to approach? Do you risk missing out on a potentially great ML Engineer in such a candidate short market? Or do you send them a message anyway and risk incurring their wrath for yet another “spammy, imprecise approach”? On top of that, how do gather accurate market information like salaries, for example?

The risk is that potential candidates are inundated with mismatched approaches. Which leads to even the decent, targeted messages, emails and calls being ignored. And if someone isn’t reading your messages they’ll never know when you do reach out with that dream job.

The ancillary organizations that support Data Science recruitment 

That’s the ATS/CRM system software providers, hiring application builders, job boards like Indeed, Monster and CV library and RecTech applications like LinkedIn. It’s a huge, multi-billion, international industry. I’m sure most of you will understand better than me how difficult this makes clustering and searching for roles and candidates. 

My workaround for this has always been to ignore job titles. Whether a business tells me they want a Data Scientist, ML Engineer or Full Stack Scientist, my first question is always “tell me what you’d like them to do”. And to the candidates that I speak to, I ask for a description of what they’ve been doing in their role. 

However, like the RecTech startups I mentioned earlier, I provide a service to help identify people and I’ve been doing it every day for 7 years. I know the complexities of this market inside out. I wholly acknowledge that an awful lot of learning has gone into my understanding. 

This has so far served me well but doesn’t help much when advertising positions or sometimes even looking through CVs or profiles. 

I can see several solutions to this problem, the most obvious would be to standardize job titles. Given the rate that the market is growing and evolving at, perhaps an industry body could step in to define titles for roles that come up in future. This would take the involvement of a large and respected body in the sector and an awful lot of time, resource, and debate!

Another is to gradually wean people onto standardized titles. Perhaps LinkedIn (being the world’s largest professional network) and/or the jobs boards could commit to limiting job title choices, by way of drop-down menus, bucketing titles into categories and gradually influencing companies and individuals to adopt their versions over time. 

A more complex solution would be to develop an NLP model that can allocate titles by scouring the web (and perhaps a knowledge graph) for details on what the person does based on previous roles and what the others with same titles in the company do. 

These solutions will need the backing of respected data organisations and industry bodies, investment in technology and adoption at grass-roots level. However, in turn, it will help people find the right jobs for them, reducing the number of irrelevant applications and approaches, help job boards and hiring apps deliver better matches, and ultimately save valuable time and resources for everyone. 

I’m sure we will get there, but the sooner we do, the more time all of us can spend on doing our parts in applying Data Science for the betterment of the world.