Data Scientists are like unicorns, they don't exist

Posting date:12 Jun 2019

Abdel Whab Turkmani is the co-founder and CTO of Util which is on a mission to democratise impact investing, existing to support an investment community that is increasingly being assessed on both its financial and non-financial practice and performance. Abdel brings experience from the Financial Services industry and founded the organisation having graduated from the University of Oxford with a degree in Computer Science and a keen interest in Machine Learning and Data Science.

He supports the idea that the term Data Scientist is both too broad and essentially, redundant.

The term Data Scientist has become muddied, a term that is so misunderstood and overused it makes me wonder if the perfect Data Scientist exists. I don’t even like to use the term anymore because actually, I have no idea what it means. It’s like a unicorn with so many required skills and specialisms that they can’t possibly exist.

How can one person be a Machine Learning Engineer, Data Engineer, Researcher and Analyst?

We are hiring at Util but we don’t hire Data Scientists. We hire for three different roles within our Data Department. Different job descriptions, different requirements and different responsibilities. The term Data Scientist is so broad that it’s impossible to find someone who can do everything but organisations seem to believe they can find someone to do all three. 

You need Data Engineers to build the data pipelines, Machine Learning Engineers to build predictive models, and depending on how advanced your technology is you might need Researchers (Research Scientists) to constantly explore new ideas, keep on top of the state-of-the-art and evaluate the applicability of new models. In some cases, you also need the Analysts who are able to look at the output of models and understand and interpret them in a commercial context. Within those roles, you also have further specialisms. For example, you have Machine Learning Engineers who are specialists in language, image, vision, sound, reinforcement learning or others.

That’s not to say that all of the individual roles set above do not intersect with one other. They integrate together tightly as anyone who joins the Data Science team will have to be cross-functional, they need to understand what everyone else does but have an edge towards one direction. Everybody has to know a little about everything but you need to know a lot about one thing.

To a certain level of sophistication, you can easily transition between the different roles but once you get to an advanced stage you need the specialty and it’s highly unlikely that one single person; or rather a single Data Scientist, is that knowledgeable.

The parallel would be something like Natural Science. At some point, perhaps 800 years or so ago, there would have been people who are “Natural Scientists” but today, as the field has developed, you can no longer be a Natural scientist. You have to be a physicist or a biologist, something more niche and specific. The same thing is now happening within the Data Science field as it continues to mature. The field has developed enough depth to merit specialists, and the search for the ultimate Data Scientists is increasingly impossible.

For full access to all of our interviews and our white paper on Identifying The Human Face of Data Science, please download a copy below.

Latest opportunities