Q&A: Anatomy of a Data Scientist
I caught up with Chris Orwa a talented and leading data scientist based out of Nairobi, Kenya. He is the Head of Data Science at Brave Venture Labs, a writer at The Dependent Variable and his blog Black Orwa, and a physical trainer. He’s also a mentor. He writes on a lot of subjects that bite his Data Science bug, practical things that people want to solve. So, taking a quantitative approach, he gathers data, analyzes it and reports on it on his blog.
It’s also worth noting that he drinks a lot of coffee
.Tell us how it all began. Growing up, did you envision yourself working in the tech world?
CO: From a very young age I wanted to be an electronics engineer. In fact, in primary school, I would take apart radios and put them back together.
In high school, I took courses that led me in the direction of computer engineering.
In college, I had the choice between computer engineering and computer science. I remember my uncle, in an effort to dissuade me from taking engineering, reasoned that if I did, I’d end up a computer repair tech, which in his eyes is undesirable.
I researched computer science and felt that it wasn’t so far off from computer engineering. I made the cut and enrolled in Kenyatta University.
How and why did you decide to specialize in data science?
CO: I was first interested in gaming development. This was after a university course in computer graphics where we worked on algorithms for rendering images.
In my 4th year, I took advanced classes in VR and AI to create intelligent game characters and make them more real. We were six in the class, one dropped.
The lecturer decided to train on data mining instead and how one could read patterns from data. He was working with banks and insurance companies, doing analysis for them. It was quite interesting – he could predict stock prices and do fraud analysis. It was fascinating how the math behind it was as complex as working on games.
Since game development is heavy and you have to work with a lot of gaming developers, which Kenya doesn’t have, I saw data mining as easy low hanging fruit. I could work alone, apply the math developed in gaming development, and the results were immediate.
So as we’d learn the algorithm in class, I went online to supplement the information with relation to data science and try to learn everything I could.
What was your first gig?
CO: My first job paying job was at a data mining company that I formed with a friend from university. We’d work for small companies like supermarkets and get paid around $500.
I then got this crazy idea – I thought if I could collect enough scratch cards, I would be able to understand the spending patterns of people, where they buy their credit from, how much they buy, etc.
So every evening, I’d scavenge the bins for these scratch cards. Once we had a good number, we sat down and wrote some code and along the way we cracked the serial numbers and learnt how airtime is distributed.
We blogged about it and it went viral. The feedback was tremendous – both from customers and potential partner brands. That is when data science felt real.
What do you enjoy most about data science?
CO: It is fresh. There is always research to do. There is no one project that is similar to the other one.
I love information – reading books, watching documentaries about different subjects – so a data science career fit my lifestyle. I get to utilize information that I learnt in class and I learn something new every day, do something new that is challenging and experiment a lot.
What skills are a must-have for a data scientist?
CO: A mix of computing and statistics (mathematics) – You will end up solving complex problems. The level of computing understanding is dependent on the problems that you need to solve. To give you some perspective, I like to kid and say to manage a large hedge fund database, you’d need a PhD in math.
Creativity aka problem modelling – You should be able to clearly define the problem e.g. we want to recruit better teachers, vs we want to recruit teachers who can teach for 6 hours with 2 breaks.
Computer programming – The more complex the problem gets, the more skilled you need to be in computer programming. Computing allows you to merge data in ways that were not thought of before – extract dates, names etc. Your level of programming also helps you to know what variables can be extracted from the data vs what can’t be extracted.
Business mindset – You first move a business problem to a data science problem then to a solution and then feed it back to the business. If you don’t understand your business case very well, you will build a model that won’t produce results.
How long would it take a novice to learn how to be a data scientist?
CO: That’s very hard to answer. It really depends on one’s learning pattern. Most of my waking hours were spent writing code.
There is a social cost to it as well. You give up hours of socializing with your friends in order to put hours into working on code and algorithms.
Begin as early as you can. I’d say you need 2 solid years of testing before you can fully execute.
Growth is mostly dependent on solving problems. I was working at iHub and got thrown into the deep end – I had to learn how to sink or swim. Most of the problems I was asked to solve were complex and I only had 3 months to solve them.
That is also how I began blogging. I wanted an outlet for all I was learning, and also a platform to showcase the different use cases and potential for growth.
How easy is to get access to data in Kenya?
CO: It is very difficult. Businesses in Kenya are very paranoid. We are a relatively small economy and data forms a significant part of a business’s capital – sharing data would be risking giving up business equity.
There is also the risk of having the data used for competitor analysis and espionage.
Relationships and trust need to exist in order for you to get access to this data. It’s not strong here, hence why I started with dumpster diving.
How is Kenya prepared for data science?
CO: We are quite far, compared to superpowers such as the US who are testing self driving vehicles. We lack the infrastructure to allow data scientists to come in and work. If you have clean data and a lot of it, all you need to do is model it. But here, you’d have to get the data, input values, do some guesswork – essentially, it becomes very difficult very fast.
Also, in Kenya, 50% of jobs are in healthcare, fintech and agriculture – nothing futuristic. As long as poverty is still an issue, we will still have people working in these fields instead of research. Data science will be back-burnered.
Click on the button below to learn more about our Mobile App & Game Development Diploma.
Here are some other resources you might find interesting.