Data scientists are among the most sought-after talents in Singapore, but the fact remains that many who are still working to be hired as data scientists are uncertain of what it really takes to become one. In fact, what exactly does a data scientist role entail?
You may be one of many people working towards a job in this growing industry, but perhaps you still have qualms about what companies look for when hiring for data-related roles, or whether data science even comprises what you’re hoping to do. What better way to find out the answers than from industry experts?
4 Reasons Aspiring Data Scientists Should Learn Python
3 Qualities Every Data Scientist Should Have, According To This Tech Recruiter
Top 3 Programming Languages You Need To Know If You Want To Be A Data Scientist
Tell us a little bit more about what you do now as well as your job experience in the tech industry.
L: I’ve been working in the tech industry for seven years and seen its evolution. I actually started out working in a bank, but I got bored and wanted to do something more relevant to my university degree in Psychology. I got into tech because I played a lot of games and developed an interest there. It’s been gratifying to see tech come to the forefront for sure, because, well, it’s the industry I chose. [laughs]
W: I help build the machine learning algorithms for Tech in Asia, with a focus on the content recommender, which personalises the newsfeed for subscribers. My passion lies in natural language processing, which is an area I believe Tech in Asia has a lot of potential in. As for my background, I was initially studying chemical engineering, but switched to Computer Science halfway. I spent a lot of time doing web development, both front- and back-end, but I also did a few side projects and was learning things that schools don’t really teach. My perspectives on data science were widened when I started working at a startup, and that led me to working at Tech in Asia.
First off, let’s clear some misconceptions about data-related roles. What are the differences between the roles of a data engineer, a data analyst, and a data scientist?
W: ‘Data scientist’ can be a very nebulous role. This is how I would break it down.
For data engineers, their job is to get data at the right state, at the right time, and in the right hands. There are three buzzwords for data engineers, these being (1) extract, (2) transform, and (3) load. They need to be able to extract the data, manipulate it into whatever state is relevant or necessary, and load it onto a site to work on.
For data analysts, they need to work with existing data and find the right answers to the right questions. Essentially, they’ll be helping people “ask questions correctly”.
For actual data scientists, they’re the ones helping people make the best decisions given the data they have. Their role involves higher-level thinking and more experimentation, meaning they’ll have to design experiments to prove or disprove hypotheses. In that respect, they have a little more autonomy. It also means that they must be excellent at data visualisation.
L: To add on, there are three major skill sets that are required for the role of a data scientist. Firstly, they’ll need to know software engineering. Secondly, they need knowledge of statistics and mathematics, since they have to craft equations and algorithms. Finally, they will need an understanding of machine learning, because they really have to know how data infrastructure works. So for instance, when problems are thrown at them, they need to know if the data is “clean” or not.
What sorts of educational qualifications do you need to work in the industry? Is it necessary to have a bachelor’s? A master’s? A PhD?
W: That’s a tough question. I don’t think there’s a single route you have to follow. I actually know someone from university who graduated with a Bachelor’s in Psychology, worked as a data analyst for a year, and just got a job as a data scientist. But he read up a lot on the fundamentals of programming and already had a strong interest in data science.
L: I believe higher education specifically in data science is not necessary or crucial, but it honestly does help. It helps due to the kind of academic foundation you gain when you do these kinds of PhDs or Masters, because you would’ve dedicated so many years in that space. However, you definitely don’t require that for a data-related role. What ultimately matters is that you do your research, that you have dedication, and that in your own free time, you’re picking up skills which can get you there.
Are jobs in the data science industry really as lucrative as they are made out to be in the media? In your experience, does the pay reflect the high demand touted?
W: [laughs] Well, I’d say yes, the pay does reflect that. The demand for people in these jobs is high, but there still is a very high bar when it comes to hiring data scientists. It shows that there’s a large gap between what education prepares an individual for, and what the job actually requires.
L: There’s definitely a demand for data scientists in the region, as well as in USA and Japan. On the whole, data scientists do get paid very well, owing to the high demand and low supply. Basic economics, you know? [laughs]
What does an entry-level data scientist job entail?
L: Well, there are no real entry-level data scientist roles exactly. If there are — that is, if they’re calling themselves that — they aren’t really “data scientists”. There are opportunities for junior positions in some companies, but in-house training is usually provided by the company such that a lot of learning is done after you get the job. It needs to be understood that data science is simply a very young discipline in the region, so it’s not a situation where Singaporean universities are training undergraduates to be able to become data scientists as soon as they enter the workforce.
What is the salary like for a bootcamp graduate?
L: It really is specific to each individual. Junior data scientists start off at around 70k per year, though I’m not sure how helpful that information is for this question, since this number tends to be for people who already have some experience in the field.
W: It’s also specific to companies. If it’s a startup or a smaller company, it might be 50-60k per year.
L: Yup. And they have to consider factors like whether the candidate is in the middle of a career switch or fresh out of school. As with any occupation, there are many variables that impact salary.
W: I think that whether you are a bootcamp graduate or not, as long as you qualify for the job and fulfill those requirements, that’s what matters.
If you want to join a company that truly needs a data scientist, what are some of the ways you can close that gap between education and industry requirements?
W: My advice is to have a portfolio and your side projects, and to document these solidly as people need to understand your projects. Employers need to be able to see your attributes through what you’ve already done. Participate actively in Kaggle, and maybe try to get to the top 300, though that is admittedly a difficult feat. Still, you should be participating actively. Other general things is to keep your LinkedIn updated, of course.
L: When a company is looking for a junior data scientist, they generally tend to look for someone with a data analyst background. In essence, this person will need to have knowledge of operational analysis, some understanding of R or Python, know technical skills, and so on.
In my own experience as a tech recruiter, I look for projects that catch my eye. I do trawl Kaggle, so as Will said, it’s important for you to have a presence there. I’ll admit that it’s quite difficult to define precisely what catches a tech recruiter’s eye, but I think if you capture all three aspects of the skill sets mentioned earlier, it raises the chances of you grabbing their attention.
If you’re really enthusiatic, you can also do the old-school method of networking at events, where you can talk about your projects and interests. I know a person from my company who went to speak to someone on our data science team at one of these events, and that turned into an interview.
W: I’d also urge you to go for industry-related events. Attending such meet-ups has the additional benefit of being able to share your ideas with others, as well as see what the industry is actually doing. Find out what frameworks and technology are relevant in the current climate so you know what you ought to focus your research on as you look for jobs.
What are some of the soft skills you look out for in candidates applying for data-related positions?
W: The job of a data scientist involves helping people make the right decisions, so it’s imperative that you have strong communication skills. You’ll need to talk to many people of various backgrounds and job functions, so you require the ability to know how to communicate what to whom. You might have to talk to the data engineers who are sorting the data out for you, you might speak to data analysts, and even business stakeholders. So aside from having an analytical mindset, it’s important for you to know how to communicate with others.
L: I absolutely agree. If you cannot articulate a complex idea to the layman person who probably has no time to listen to you explain technical data, you’ve failed at your job as a data scientist. Your role requires you to be able to persuade people with data, so you can communicate what needs to be solved convincingly. That’s why communication skills are so crucial.
What questions do you usually ask to see if the candidates possess such skills?
L: I like to ask the candidate, “What is the one product or project you are most proud of and why?” This is so I can figure out whether it was their own solution, whether they were the one who saw the issue, whether they were proactive enough to step up, wrangle the team together and actually solve the issue, and how much money they saved or made for the company. There’s a lot you can gather from their answer by finding out the rationale behind the project they say they are proud of. For me, that rationale helps me see the qualities of the person and whether or not they would make a successful data scientist.
How should one go about preparing for technical interviews?
L: Sometimes technical skills required are unique to each company, so it’s hard to say. But there’s a lot of information out there on Google and platforms like Reddit, so do your research.
W: Agreed. You can see what questions in particular people have failed to answer in these technical interviews, and attempt to figure out the answers yourselves. I’ll always remember one question that stumped me when I was being interviewed years ago for a data scientist role. I was asked, “Can you name me some loss functions and what loss functions are good for?” And I didn’t really know what situations to use these models for! But it helped me realise that recruiters want to know whether the candidate genuinely understands the models they’ve learnt and where certain models can perform the best in given situations.
Given that it is still a relatively young industry, how do you see the data science industry and data-related roles evolving over time? How have you seen it evolve over the years?
W: I remember in 2015, I was in San Francisco doing an internship. Back then, there were no data scientist, data engineering, or data analyst roles. There were only not-so-visible research jobs that used to be back-end roles. I’d say that the future holds more diversification and more specialisation, such that the lines between each role will become increasingly pronounced.
L: I agree. In the future, or even perhaps right now, there’s going to be not only more specialisation, but the industry will gradually understand how to scale in each specialisation.
What industries have the highest demand for data scientists?
W: The transportation industry, as we know, is booming. Companies like Grab and Uber are picking up speed.
L: Healthcare is another industry that I believe is going to need a lot of help transitioning from more traditional data architecture to newer technology.