👨🏼‍💻 An Interview with Shane van Heerden

Life in the shoes of a Senior Data Scientist at Luno

1. What is your name and role and Luno?

Hi, I’m Shane and I’m a Senior Data Scientist at Luno.

2. What has been your career journey?

After finishing school, I, like many others, was a bit stumped about what career I wanted to pursue. At the time, the best I could do was lean into the subjects I did well in and was most passionate about at the time, namely: Physics, Business Studies and Maths. I believe this is what ultimately convinced me to pursue an undergraduate in Industrial Engineering at Stellenbosch University in 2013, a degree that ultimately taught me to think like an engineer with a business hat on.

Throughout my undergraduate years, I discovered my deepest passions ignited when a module incorporated elements of programming and mathematics. A pivotal point in my career journey occurred right at the end of my undergraduate in 2016 from an unexpected source: YouTube. A recommended Stanford lecture introduced me to the realm of “Machine Learning,” a paradigm enabling computers to make intelligent decisions without explicit, programmed instructions. I was hooked! So much so that I decided to pursue a PhD in Data Science and Machine Learning which I successfully defended at the beginning of 2020.

Transitioning from academia, I was eager to discover how my newfound, mostly theoretical Data Science and Machine Learning knowledge could actually solve real-world problems. My first job at a Data Science consulting startup gave me the chance to do just that. I worked on a bunch of different Data Science projects in various industries, which was a great learning experience. I think this experience really prepared me well for the unique problems I would encounter at Luno.

I have been working at Luno for about 2 years now and I’m amazed at the personal and professional growth journey I have been on. During this time, I have transitioned from being more in the fun-loving Customer Service space to now grappling with the complex world of anti-money laundering and fraud prevention. Each day, I find myself even more amazed at the extraordinary lengths Luno goes to make sure customers are safe.

3. What do you enjoy about your role?

The best thing about being a Data Scientist is the variety and depth of the problems that I get to work on everyday. Coming from a strong research background, I enjoy the intellectual challenge of trying to crack a really tough problem. Even more than that, the satisfaction I get from seeing that solution driving real value is what gets me out of bed every morning.

4. What are the top three tasks you would complete day-to-day?

One of the greatest privileges about being a Data Scientist is that your skill set gives you the ability to solve many different types of problems in many interesting ways. In that same breath, no two days ever look the same for me. Some themes of tasks do remain the same though:

Most mornings start off with connecting with my fellow Data Science colleagues in our morning standup. This is usually followed by an informal collaboration session where we can exchange ideas or help each other out if we are struggling with a specific problem. I’ve found having this space to connect is so important especially in this more digital work world we now find ourselves in.
If I’m at the very beginnings of understanding a problem and what possible solutions I am considering, most of my time will be spent engaging in meetings with stakeholders or engineers, reading blog posts on how other people have approached similar problems, searching for data to see if it even exists, or formulating my thoughts on a blank canvas with sticky notes and diagrams so that I can easily bounce some of my ideas off my other colleagues.
If I’m in the later stages of a project where I’ve already converged on a solution, most of the time I’ll be building out a data pipeline to get the data I need into a usable form, tinkering around with specific model configurations in a Jupyter notebook, thinking about the best way to convey a piece of data or preparing for an upcoming presentation to stakeholders.

5. What are the top skills you need to master to succeed in your role?

This is a tough one. Naturally, as a Data Scientist, one may gravitate towards those hard skills like Python, Statistics or even Deep Learning. Although important, I have found that what really distinguishes someone in this field is their ability to pair their hard skills with some very important soft skills. Here are the top 3 skills I think every Data Scientist should master:

Have a curious mind. The field of Data Science is constantly changing. People around the world are discovering new and interesting ways to solve problems using Data Science techniques. An approach that was state-of-the-art a month ago will soon become obsolete. Although I would definitely not advise going down every new rabbit hole, I believe remaining deeply curious about the Data Science field is incredibly important for fueling new creative ideas to problem solving. Moreover, let this curiosity extend into the way in which you understand and approach people and problems!
Learn how to listen. One of the common pitfalls Data Scientists very often stumble into is rushing into a solution without truly understanding the problem they are trying to solve. It takes a great deal of humility to put all the fancy models and algorithms aside to truly listen and understand the problem a stakeholder is facing. Encourage open dialogue by using prompts such as “Could you share more about your process?” and “Can you walk me through your plan once I provide you a solution?”. Listening ultimately nurtures a stronger connection with the problem itself rather than being fixated on a preconceived solution.
Master the art of storytelling. One of the early lessons I learnt in my Data Science journey was that no matter how ingenious your solution is for someone’s problem, it won’t count for much if you can’t explain it in a way that the decision-makers can easily grasp. Slapping a bunch of graphs onto a PowerPoint without considering the journey you’re guiding stakeholders through is a recipe for disappointment. Our brains are just not good at retaining raw facts for very long, and they often don’t drive the meaningful actions we’re aiming for. Rather, we are wired to understand, retain and deeply connect with people through stories. Take your stakeholders on the journey of how you arrived at your solution. Mastering storytelling, of course, takes practice, and I’m definitely still learning. But one thing you can do is seize every opportunity you get to present, even though it may seem scary at times!

6. What advice would you give to someone looking to develop into the position you are in?

Leaning very much on the the skills I previously mentioned, there are a few nuggets of advice I would have loved to give myself earlier on my journey as a Data Scientist:

Don’t focus too much on the method. Too often, as a young Data Scientist, one can get caught up in debates like “Should I use Python or R, Plotly or Seaborn, PyTorch or Tensorflow?”. Just pick one! What’s more important is that you are able to extract value from data using one of these tools.
Plan your problem approach. When embarking on tackling a Data Science problem, there is often no shortage of possible solution approaches you could begin to explore. If only you also had copious amounts of time to go down each rabbit hole! Through experience, you’ll get better at predicting which solution avenues are likely to yield the best results and knowing when a path is likely heading towards a dead end. Taking some time ahead to plan out what you are thinking of exploring and validating or even improving on some of these ideas by getting feedback from other colleagues and/or stakeholders can be super crucial to your success.
Understand your solution deeply. I know it may sound obvious, but once you have proposed a solution to a problem, make sure you understand your solution deeply. If, for example, you have proposed a new model that is going to predict fraud, make sure it’s trustworthy by knowing how that model is coming to its conclusions. Don’t only look at metrics like AUC or R² but rather interrogate your model by looking at how it misclassified specific examples and understanding its shortfalls. In this process, it’s also important to document how you arrived at your solution and all the assumptions you may have had to make along the way. Ultimately, this will build trust in your solution among your stakeholders which will hopefully lead to the desired impact you are wanting to achieve.

7. What has been your highlight at Luno?

The crypto landscape is constantly changing and filled with new challenges. Being in a company with the enthusiasm and willingness for innovation makes it so easy to get excited for work every morning. It may sound cliche, but this is why the people at Luno are an everyday highlight! I am truly grateful that I have the opportunity to learn from such incredibly talented colleagues that push me to become the best version of myself. Moreover, I’ve been fortunate enough to have an exceptional manager who not only fosters an environment where I can develop the skills I want to grow in but also trusts my opinions and views me as a partner in Data Science discovery.