.svg){kind=small}
“In the world of machine learning we see everyday, especially in the last few years, that more and more systems and real-world use cases use machine learning. Not just for recommendations in music, but for very impactful purposes which is very exciting.”
Another week, another interview. Once again we get to meet a Hopsworks team member, this time we’re introduced to Software Engineer Antonios Kouzoupis. Antonios talks about his background in Academia and what part about his work with feature stores he finds most exciting.
Antonios: My name is Antonios and I’m originally from Athens, Greece. I’ve been working at Hopsworks since day one, so basically since early 2016/2017. Before that I was working as a Research Engineer at the Research Institute together with Jim Dowling, our CEO, and that's how we met.
I came to Sweden to do my Masters Degree in Distributed Systems at KTH and I had Jim as a professor in one of my courses. Later on he was also a supervisor for my master thesis. So there are a lot of people at Hopsworks who have a background in Academia, either with a Masters degree or a PHD, who are attracted to the work we’re doing here.
Antonios: My bachelor's degree was in computer science and like I mentioned before, I did my masters in distributed systems. I kind of always had an interest in large scale computer systems and I was always amused by how computers interact with each other and how they kind of worked towards a common result. Hopsworks is a very complicated platform and system and we try to hide that as much as possible from the end users. That part has always been really exciting for me, both from the start and today. In the world of machine learning we see everyday, especially in the last few years, that more and more systems and real-world use cases use machine learning. Not just for recommendations in music, but for very impactful purposes which is also very exciting.
Antonios: MLOps is probably the most popular field in computer science right now. So it's an area that's constantly evolving. On one hand you cannot settle, but on the other hand you see things constantly changing. You have to keep up with all the new technologies and you have to question yourself and what you have already learned through your studies. It's also very interesting that what you have studied, you can apply practically in a field that actually has an impact and you feel like you’re doing something useful.
Antonios: Most of my sources of information are different articles online, for example on Hacker News or Twitter. But I want to give a word of advice. Something that I've seen more frequently, in particular from our interaction with different customers, is a bit more fundamental ideas about feature stores and how to model your data to actually make impactful model training. This is where I see a lot of people struggling and it's the first step that you have to get right, because after that it's quite difficult to move backwards and start changing your data architecture and your machine learning operation. So something that I would like people to consider is when you move from modeling your Machine Learning System from your laptop into production, just think about all the different aspects and all the complexity that will come further down the line.
Listen to the full episode:
.svg)
