5 questions with Jakub Nalepa
Our guest is Jakub Nalepa, Head of Artificial Intelligence department at KP Labs, Associate Professor at Silesian University of Technology, and Machine Learning Architect at Graylight Imaging
Jakub answers Ewa’s 5 insightful questions.
– what encouraged him to choose his career path
– what he has learned from his mistakes in the past
– what advice he would give to his younger self
– what benefits and opportunities data and its analysis have to offer
– what are the advantages of peer review in the world of science and business.
Ewa Banaś (EB): Hi, I’m Ewa. Welcome to our new IT Insight series, Five Questions. In this series we’ll discuss our speakers’ experiences and lessons learned today. My guest is Jakub Nalepa. Great to have you here.
Jakub Nalepa (JN): Absolutely, my pleasure.
EB: Before we move on to the questions, could you introduce yourself a little bit to our audience?
JN: Yeah, sure. So maybe a long story short, I got PhD in 2016 in machine learning and after that I actually joined the university and I tried to bridge both worlds scientific and industrial. I’m an associate professor at Silesian University of Technology and I also work at KP Labs where I am head of AI department trying to oversee all activities that are concerned with AI, in for instance remote sensing. But also I’ve been doing medical imaging for years already, almost 15 ish and that is also an important part of my work.
EB: Thank you for that intro. I’m really looking forward to our talk. So let’s begin. You have a vast experience in machine learning and data analysis, but if you aren’t doing this, I mean being a machine learning architect and associate professor, what would you be doing career-wise?
JN: Right, so you’re launching difficult questions for start.
EB: That’s my task here.
JN: So I think I must say that I always enjoyed solving problems and I had a very early start during my studies I started working on data analytics and I saw a great potential in the data that we acquire every year. So even 15 years ago we observed that there is huge amount of data being acquired and stored. So I wanted to make sure that we can benefit from that. And I think if I were not to be a machine learning person, I would probably go with medical imaging field or medicine in more general because the problems that I tackle every day are, in my opinion important. So I try to bring value to the end user. I want to make sure that we can benefit from the data we capture. But on the other hand it is nice to see that things are impactful so we can make change in different fields. So that could be for instance medicine, it could be geoscience or remote sensing. I think I would just go for solving difficult problems in the best possible scenario for a good reason. And this good reason is to have impact.
EB: From what you are saying all I can hear is data, data somewhere there in the background?
JN: Yeah, absolutely. And we gather data but we need to get knowledge from this data.
EB: Definitely. And when you look at your career, what advice would you give to that 20 years old you?
JN: I can say that I think I’m fortunate enough to say that I really enjoyed my career because again that was a very, very early start. I started working on this during my studies and I think science is the art of asking good questions. You need to have a good question to dig into your problem that you want to understand. And I think the most important advice would be to ask questions and not to be afraid of asking questions. Even if you talk with more experienced people, because that is where you learn and you need to learn from experience. So why not to exploit the knowledge that’s already in more experienced and more senior people in your field. So I think asking questions, not being afraid of asking questions is from my point of view, super important. And it can make change.
EB: It all starts from the good questions, doesn’t it?
JN: Absolutely. Yeah it does.
EB: Yeah. We all learn by our mistakes and I’m curious if there was a moment in your career that where, when you made a mistake and you learnt something really important out of it?
JN: Lots of them. I think every experiment could be successful or not successful and if we do not succeed with a single experiment, we can learn from our experience. And if we for instance, have a hypothesis on some phenomena that we want to verify, we can get negative result of our experiment. And that is also very important because we can again learn from this and we can try to fine tune our approach to tackle the problem in an appropriate way. So there are lots of problems that you might tackle every year, every day and during your career, but you need to learn from them. So yeah, I think that is a super important part of our learning experience to be able to benefit from your failures and the failures come and that is something which is absolutely normal for junior and for senior people as well. The question is how to benefit from this.
EB: Exactly. And have you got any particular example in mind that you could share?
JN: For the failure, I think, I won’t say it is a failure as such, but I think that might be an example from the project that we had recently. We were designing the algorithm for segmenting lesions in brain in magnetic resonance imaging of brain of children. And the data was very difficult, which means that it was very imbalanced. The quota of the data was not very nice and we need to somehow resolve that problem. And for the very start of the project, we basically failed, we had the algorithm that worked for adults and it totally failed for children. It is good because we always want to understand why things fail. And if you understand this why they failed, you can fine tune or redesign your approach to better solve this problem. So that was a matter of fine tuning or updating the technique to a specific problem. And at the end of the day we succeeded and we do have the algorithm that works for that kind of difficult data.
EB: That’s a really great example. Thank you for sharing this one.
JN: Yeah, there were some situations where we had the data of low quality and that was the most important part of the entire project to make sure that we are in a very good starting point to move forward. Because afterwards we can, at the end of the day we learn from data. If the data is low quality, then it is likely that your technique would be of low quality as well.
EB: Definitely. And I think that those checkpoints are very important in data. But generally in life, make a hard stop and think-
JN: Absolutely, yeah. And also to track because you can quantify the progress of your project. Even if we look at the scientific projects, it is perfectly fine to track. You should do that to make sure that you know where you are with your project and with your ideas.
EB: Definitely. And set goals so that you know where you’re going.
JN: If we can quantify them, we are in a perfect place.
EB: Definitely. Looking at the global situation and how it may influence IT industry, what challenges you predict you’ll have to tackle within your future?
JN: There are lots of challenges in different fields. Again, we acquire a lot of data, so the amount of data is enormous. So every single second we gather more data, which is difficult. And this difficulty comes from the fact that we capture the data in a continuous way. The data might be multi-model, which means that we have different features of data and different characteristics of the data. The quality of the data might be different. So we cannot assume that we have good data and we can get knowledge from this. And I think making sure that we can benefit from those heterogeneous data in a particular field. For instance, in medicine we know that Covid 19 outbreak was a huge thing worldwide. So we need to be prepared for the future because it’s likely that we need to face that problems in the nearest future as well.
So we need to know how to utilize the data that we have because we know that we have lots of data, but this data should be utilized in a good way to make sure that we have robust and correct answers coming from data. And medicine is one example, but on the other hand we may look at remote sensing where you can capture images on board a satellite, which means that you can cover a lot of information of our globe, of the world. But the question is how to get the knowledge from raw pixels. Because the pixels are just the data. The question is how the user, the end user could benefit from this. Could we have better agriculture? Could we have tracking processes across the globe? And that kind of, I think questions are very important and they are getting even more important nowadays.
EB: Definitely, we see it by our own eyes right now. And I know that the academic context is very close to your heart being an associate professor. I’m wondering how you perceive conducting research and doing business. What’s the alignment there?
JN: I think bridging those two worlds is important because in the scientific world we publish. So we design the algorithm, we verify it, we validate it, and then afterwards we publish. And this process of publishing is important because when you submit a paper, for instance, to a conference or to the journal, you know that other experts in the field would actually look at this paper and verify if it’s correct, if it makes sense. So we have this peer review process, which means that at the end of the day, if you have the paper published that was already reviewed, which means that it is a good evidence that your things work in practice. And that is something which is also important in industry because you want to make sure that things work, but you also need to be able to prove that your things work correctly over the data you expect. For instance, in medicine it’s again very important to make sure that you have your techniques already validated because they may also impact patients.
EB: Especially in medicine.
JN: Yeah.
EB: I guess, yeah. Thank you for our talk today for sharing your experiences with us and with the audience and thank you, our listeners for listening to yet another IT Insights episode. Thank you and see you soon.
Check similar insights
