Mantas Lukauskas, a Data Scientist at Zyro, is one of the brains behind Zyro’s AI-powered tools.
When he’s not working on teaching the AI to learn new things, he’s pursuing his academic career in machine learning as a Ph.D. informatics student at the Kaunas University of Technology, aspiring to get his work published in scientific journals.
In this article, he sheds light on how his interest and talent in exact sciences led him to do something he always said he wouldn’t do – in other words, programming.
From a math protege to an economics student
I have always done very well in the field of exact sciences. It was evident from when I was about 4 years old: I could not only count from 1 to 100, but I was also able to multiply, add, divide, and so on.
It’s hard to pinpoint what the appeal to math and numbers was originally, but in general, I have always liked accuracy. There’s no better feeling than following exact facts and figures and drawing conclusions based on them.
While it made sense to go into university to study IT or math, I never wanted to become a programmer. And the aspect of becoming a math teacher didn’t appeal to me either, so I ended up applying and getting into Kaunas University to study material science and nanotechnologies.
During my undergraduate degree, I also studied related economics and ended up graduating with an honors degree in 2017. But I felt like it wasn’t enough – I wanted more. It really hit home while I was writing my bachelor thesis: I realized that I didn’t want to just sit in a lab and do theoretical research all day. I wanted to get my hands dirty and do something more practical, with direct real-life implications.
How I ended up at Zyro
I hadn’t even finished my studies when the opportunity for paid internships arose. An automotive sensor production company allowed me to gain experience and earn some cash while I was finishing up my undergraduate degree.
I ended up staying on for more than 3 years as an Associate Product Manager and a Product Management Analyst, as I was able to continue working there during my master’s degree.
One of the things I’m most proud of is that I was able to juggle work and my studies well. In fact, I ended up winning 8 university talent scholarships during my studies, scored top grades, and in 2019 I won the Confederation of Lithuania industrialists Petras Vileišis award.
I told him that I was, but wasn’t able to talk about the details until around 1.5 months later. He agreed to wait, and before I knew it, I was working as the newest addition to Zyro’s AI team.
Understanding the quality and value of your data
As a Zyro Data Scientist, I gather data, then clean it and use it to train different AI models to do different things. Our team also works on testing and iterating the models after they have been deployed and are in production.
The most important thing we have to pay attention to, in my opinion, is the quality of the data we use. That’s because to get good and reliable insights, you need high-quality data.
Also, we have to be mindful of the speed at which we do things: nowadays you can’t wait around and take too long to analyze your data. Otherwise, someone else might beat you to it, and your work has been for nothing.
A good Data Scientist also needs to know the value of their data. That means that even before you begin your analysis, you need to know what you want to achieve.
If you’re working on clustering analysis, you are keeping an eye out for particular groups of users; or, if you’re working on data classification, you want to create or choose a good model and later use it to effectively classify your data.
Deep learning isn’t just for big companies
One of the biggest myths I come across time and time again is that small and medium-sized businesses can’t work with deep learning. Sure, it’s faster to train effective deep learning models when you have access to powerful hardware. But you don’t need a supercomputer to get started: my first deep learning model was made with an old Nvidia GTX 960 (2GB).
A lot of people also seem to think that you just need to create a better model and you’re done. But many seem to not factor in that the better and more thorough your data collection, cleaning, and differentiation processes are, the better your outcome will be overall.
80% of your work should be focused on the data you’re using – the model and model deployment are just a part of a good AI application process.
At Zyro, I can combine my Ph.D. and my work
I don’t think I could ever go back to working for a traditional company after Zyro. The best thing about working here is the values-based company culture and the emphasis on freedom and responsibility – meaning that I can combine my Ph.D. studies with my work.
As I don’t like to depend on other people to do my job well, Zyro has enabled me to plan my work the way I like it. Thanks to the open communication culture, everyone is encouraged to present their ideas and own what they do – both when things go well, and when they don’t work out.
The fact that you can choose to work remotely is also a huge advantage for me since I have my own dedicated home office already. And since everyone works to the highest standards, I feel like I fit in perfectly since I’ve always strived for the best outcome, be it my work or my studies.