Meet Sofia, she can change the way you shop, think and behave using advanced machine learning and artificial intelligence.
Sofia Hörberg left the role as a data scientist at eBay to lead the data science initiative at Combine – South in November 2017. At Combine she found the possibility to build a new team from scratch and she now has the freedom to work with all interesting customers and projects, ranging from automotive to medical applications.
Everybody is talking about Data Science today, but you already have numerous years of practical experience in the field. How come you were so early in the game?
During my studies over 10 years ago, I heard about a startup that used biology and genetics to solve advanced problems and it sparked an interest that ultimately changed the course of my academic life. I had finally found a way to combine my interest in biology, mathematics and programming. After studying all relevant courses, I ended up doing my master thesis using machine learning to predict outcome after heart stop.
Even now days it’s difficult for data scientists to get real life experience after studying, how was it for you?
I was lucky to get a job at a start-up that had good connections with companies in Silicon Valley. They were very curious and eager to start working with machine learning and AI long before it was the buzzword of the day in the rest of the world.
So, what was it like getting out in the industry?
Working with large American retailers and online e-commerce companies gave us access to big amount of high quality data. This was of course a dream for a data scientist and gave me and the team possibilities to develop and experiment with algorithms for real-time machine learning. We quickly learned that the academic research in the field entirely lacked focus on real-time events. The main difference with real-time systems compared to working with stored data is that in real time systems you actually have the possibility to instantly affect the outcome of your algorithms. For example, you can give better product recommendations where you increase sales and learn more about the user.
What type of challenges did you encounter?
In order to set up a real-time system that could train machine learning models efficiently, you need a very efficient infrastructure that can process big data extremely fast. Since there were no existing software that fulfilled our needs we started to develop our own tool, the Expertmaker Accelerator. I was a little bit surprised how much fun it was to work with extreme optimization since I’m not originally a computer nerd. But it went really well and it was very satisfying when you can speed up your algorithm by magnitudes.
That sounds awesome! It must have generated some buzz in the industry?
Well, I’m not sure how famous the tool was outside our company. But the customers were very satisfied, and we were bought by eBay, which was one of our customers. The tools are now open sourced by eBay and free for anyone to use, which is great. That means that we can continue to use it to deliver efficient machine learning for big data projects.
So why did you change from eBay to Combine?
I felt it was time to take on a bigger responsibility and not only developing the technology aspects of machine learning, but also how and where data science is used. Combine gives me the freedom to do just that. I am also able to work with many different companies, where I have full freedom to choose who I collaborate with. I firmly believe that you become a better data scientist if you are exposed to diverse challenges. I talked to many different companies, both locally and abroad, but eventually I choose Combine because of the technical level and start-up like company culture that I really liked.
What’s your vision for Data Science at Combine?
To to create real value for our customers. This is an area with a lot of hype, and many enthusiastic developers. It’s very easy to start cutting corners on the engineering quality and the scientific soundness and you end up with useless solutions. To avoid that we work a lot with mentoring and peer reviews. I also think it is important that data scientists are good developers as well and we make sure to only hire people with really good programming skills. Our work consists of a mix of on-site consultancy and in-house projects, so we can get as much experience as possible and then share it with our team members. This means that we can solve all our customers needs and even become their complete data science department, taking responsibility of all aspects ranging from storing data securely to developing machine learning algorithms and hosting live production solutions. Among other things, this will enable companies without internal data science departments to quickly and cost-efficiently jump on the data science train.
Read more about Sofia and her teams work with the Expertmaker Accelerator on eBay’s tech blog:
For more information about how Combine can assist you with data science expertise, contact Sofia Hörberg.