How do you build an AI system that gives doctors👩⚕️advice on critical decisions in such a way that they actually trust it? 🤔This is the obstacle that Andreea faced when building an algorithm for detecting contractions coming from the endometrium, which is a muscle on the belly of mothers, in order to determine risk of preterm delivery.
Andreea realized that by involving doctors, nurses, and other domain experts in the process 🔁 of building the algorithm, she could gain their trust. This process of co-creation led to a more accurate, robust, explainable, and ultimately trustworthy AI system 😇✅. In her (slightly paraphrased) words 💬:
“The doctors were totally involved in all the models and algorithms that were integrated into this platform. What I noticed was that the doctor who was working closest with me was totally passionate about this—really into it. I think that’s the key message here. You need to have the domain expert involved in the co-creation of it.
More broadly, a lot of people don’t trust ChatGPT. Did you know ChatGPT has been banned in a couple of corporations in the Netherlands, and I know for sure in Europe as well? Their employees were literally uploading sensitive data of their clients to summarize and fast-track their work, and it’s just not trustworthy yet. But I do believe that for these large language models to be trusted, they will need to be developed in-house. A lot of companies are already working on developing in-house large language models with their own data, documents, or whatever, and keeping it in-house—hosted virtually and securely—so they don’t have to use models hosted in some unknown cloud with unknown access.”
In the latest episode of the UserExperience.org podcast🎙️, I sat down with Andreea and we chatted about her career in tech, her thoughts on the future of AI, and the overarching role that design plays in the world. To listen to the episodes, click here:
The text has been slightly edited using ChatGPT 4o for readability.
Olivier: Welcome to another episode of the userexperience.org podcast. With me today is Andreea Moga, and she is a serial entrepreneur. She started out with a master’s degree in Electrical and Electronics Engineering, also a master’s degree in techniques of processing signals, which I’m actually quite interested to hear more about. Over the years, she started a bunch of different initiatives, including Tech Labs Rotterdam, where she’s the Chairwoman. I’ll be taking part in that in March, so I’m excited about that. She’s also a tech mentor, helping to empower people and get the potential out of them. She’s also a co-founder of the Rotterdam Tech Social, which I guess is a sort of community of peer-to-peer mentoring, and a whole bunch of other stuff. Also working right now part-time at SURF on EdTech, related to AI, yeah?
Andreea: That’s actually more supercomputing and high-performance computing.
Olivier: Cool. Also, I guess related to the workshop you attended yesterday?
Andreea: Yes exactly. Thank you for inviting me here. I’m really excited about it.
Olivier: You’re welcome, so we actually met quite a while back, I think almost three or four years ago, something like that, during the coronavirus pandemic. We were both in a kind of, not a book club, but a mentoring group online related to entrepreneurship—kind of just startup founders, you know, helping keep each other accountable and discussing. Then we kind of lost contact for a bit, and then we ended up sitting next to each other at The Next Web conference last year. And then you were like, “Hey, I think I recognize you.”
Andreea: Yeah, actually, we met online for the first time, if I remember correctly. So that was like, you know, after a couple of years, seeing each other in real life, it was like, “Wow, it’s like meeting online,” because everything was like that during the pandemics. People who met for the first time, it was pretty much happening online, so it’s quite funny.
Olivier: And yeah, today we’re here to talk about your career, how you got to where you’re at now, and what you’re passionate about. Since this is the userexperience.org podcast, I’m also interested to hear about how design, product design, has shaped what you’re into. But before we get into that, maybe let’s just take a step back to where you started. So you did a bunch of things. First, you started with your bachelor’s degree in Electrical and Communications Engineering. Could you tell me a little bit about what motivated you to do that?
Andreea: So it was actually electronics and telecommunications in Romania. I’m originally from from a small hometown in the eastern part of Romania. I studied in Iași, which is also on the east side of Romania, for my bachelor’s and master’s as well.
During my master’s studies, I got an Erasmus scholarship to work on my master’s thesis here at Eindhoven University of Technology, where I focused on a biomedical engineering topic. My major was called Modern Techniques of Processing Signals. Nowadays, we would probably call it data science—anything that has to do with data processing. So that’s what happened. I came to Eindhoven.
Olivier: What was your thesis topic about if you remember?
Andreea: Yeah, I remember. That was in 2013, I think, so that’s like 11 years ago. The topic was building an algorithm for detecting the contractions coming from the endometrium, which is a muscle on the belly of mothers at risk of preterm delivery.
There were a bunch of signals and data being recorded from electrodes placed on the belly of the mother. The study was done in a hospital under certain protocols by doctors, gynecologists, and others. I already had the data, so I didn’t have to do the study myself. I worked together with a gynecologist, meeting every two weeks, to build that algorithm and integrate other algorithms that were already in the department back then at the University of Eindhoven.
Olivier: Was that using AI and machine learning? I know these are buzzwords now, but was that how it was working?
Andreea: So that was really using mathematics and processing and sampling all kinds of data points that were coming from these recording devices. We were designing a geometrical model, if I remember correctly. I was measuring the velocity, which is the speed of how the electrical signal actually goes from one electrode to another. By filtering at certain thresholds, certain frequencies, you can detect if that’s the endometrium, which is the layer of muscle we were interested in. So, yeah, I could say that nowadays we would call it data science. Of course, AI is also split into certain complexities—there’s the basic one, machine learning, and then there’s deep learning, and then further on, generative and interactive AI.
Olivier: And then back to your thesis—did they end up using that in their actual work?
Andreea: Yes. Next to creating the algorithm, I also built a graphical user interface, totally in MATLAB, which is a very old-fashioned tool, but you can build graphical user interfaces in MATLAB. I integrated a bunch of other algorithms created by other people in the same department, and next to that, I integrated my own algorithm and after I left someone else took over and took the project further.
Olivier: Cool. And how did people react? Did you also talk to users, like doctors? Who was the target audience of your application?
Andreea: That’s a really good question. I have to confess, the user experience was not really a focus—the focus was on creating the algorithm and integrating the algorithms. The main UX input that I received was from the doctor who was coming every two weeks to give updates and feedback on how the interface looked. But it wasn’t just about the interface; it was also about how the algorithm was working. Does it make sense how we’re calculating and modeling things? We were really modeling a geometrical model of electrical signals.
Olivier: So I guess the input would be those signals coming from the pads. What was the output?
Andreea: The output was really just bits of signal. I remember we had like five signals coming from these electrodes, and I was supposed to detect those windows of prediction—the contractions that were most likely to be labeled as having a risk of preterm delivery. It was really just to capture those windows in the signal and annotate them.
Olivier: I’m trying to visualize what the doctor would see on the screen.
Andreea: I think I have a screenshot of the interface on my LinkedIn if you want to check.
Andreea: What I want to share—something important from my experience back then—is that I had never built a graphical user interface in MATLAB before I started working on this project. Before coming to Eindhoven University, I was so impressed by my own strengths and powers because I had never done something like that before. For me, it was a moment of enlightenment, but also a realization of how much you can achieve, especially when you’re in the right environment. There were more people working on similar projects, passionate people in the department. I think being in that environment, working together with passionate people, sharing ideas, sharing feedback, honest feedback, and empowering each other was so important. That’s something I really received from the department. Maybe as advice for any younger students or people working on their thesis, it’s really important—the environment you’re in will affect the performance of your thesis, as well as the people you surround yourself with while you’re working on it. Of course, if you don’t like it, you can change it or create your own environment and surround yourself with people who could help you achieve that.
Olivier: Yeah, cool. We’ll talk more about that, about shaping your environment. Maybe one last point about this before we move on—I’m really into Explainable AI right now, and the whole idea of how people interact with AI, how they trust it. Could you tell me a little bit more about that in terms of this project? So, I guess the doctor sees the output of this AI system. Did he trust it, or how was that?
Andreea: Yeah, so I have to share that doctors were totally involved in all the models and algorithms that were integrated into this platform, as far as I know. The doctors were totally involved in all the models and algorithms that were integrated into this platform. What I noticed was that the doctor who was working closest with me was totally passionate about this—really into it. I think that’s the key message here. You need to have the domain expert involved in the co-creation of it.
Olivier: Do you, though? Because with something like ChatGPT, you know, it’s just a bunch of data thrown in there.
Andreea: A lot of people don’t trust ChatGPT, that’s true. Did you know ChatGPT has been banned in a couple of corporations in the Netherlands, and I know for sure in Europe as well? Their employees were literally uploading sensitive data of their clients to summarize and fast-track their work, and it’s just not trustworthy yet. But I do believe that for these large language models to be trusted, they will need to be developed in-house. A lot of companies are already working on developing in-house large language models with their own data, documents, or whatever, and keeping it in-house—hosted virtually and securely—so they don’t have to use models hosted in some unknown cloud with unknown access.
Olivier: Yeah, okay, cool. We’ll definitely get into more of Explainable AI and stuff. So after your bachelor’s thesis, you ended up doing two master’s degrees. Could you tell me a little bit about your thought process of why you chose those?
Andreea: Let me just correct you on that. It’s not really two master’s degrees. I had a master’s degree from Romania, and while I was a student in Romania, I got an Erasmus scholarship to do my master’s thesis at Eindhoven University of Technology. So I’m really grateful that there was an opportunity to apply for this Erasmus scholarship. I think I was the first student between my university and Eindhoven University to have an Erasmus agreement. I’m happy that I could help kickstart this collaboration because I know that multiple students after me were able to apply more easily since there was an inter-institutional agreement that I had to take care of in the beginning as well. I’m happy that I was able to help bring up relationships between institutions.
Olivier: I think that’s a theme in your life—paving the way for people, inspiring people, and getting people together. I remember when I was at the Tech Labs in Rotterdam, you were like, “Hey, Olivier, have you met this person yet?” That’s a really nice environment to be in, where everyone’s helping each other out and looking out for each other. So that’s cool.
Andreea: Yeah, I think that’s something specific to Rotterdam as well. Really, people like to connect and help each other. That’s something I don’t see much in Amsterdam, I have to say.
Olivier: Okay, cool. Then moving on, I’m taking a look here, and you mentioned that you were a software engineer at the European Space Agency. Would you like to talk about that?
Andreea: It was a hackathon—48 hours. That was during my software engineering life. After I graduated with my master’s thesis, I started as a software engineer at a big consulting company, Canadian, quite global. I worked on projects dealing with satellite data, specifically navigation data from European navigation satellites and GPS satellites. The one that you’re seeing there at ESA, that was also one of the projects.
I think that’s one nice thing about working in tech while being a student in tech—you can sign up for hackathons, even if you don’t have a team. You can just show up, and they’ll team you up with other people in the room. It gives you great experience to learn and use data and see what companies are interested in. This was organized by ESA, the big European Space Agency business incubator. It was also an opportunity to network and engage your creative skills. I remember what we thought about was using data from Earth observation satellites.
Okay, so they use data from navigation satellites, but Earth observation satellites are actually satellites in orbit with sensors on board—mostly to get computer vision data, pretty much images taken all around the Earth. You have to wait for a couple of cycles for the satellite to go around the Earth until you can create an entire picture of the Earth. These are sometimes used for meteorological purposes as well. Organizations like KNMI, you know it?
Olivier: I got an alert yesterday. There was a storm in the Netherlands.
Andreea: Oh yes, the storm! Were you okay?
Olivier: My window made a lot of noise, but yeah, I was completely okay.
Andreea: So yeah, basically, what we had to do was use that data in 48 hours. To be honest, in that time, you can’t really build a prototype. We only had time to think of the business case—what problem are we solving? And the problem was… actually, what we thought about building back then was an Earth Care app, which ties into the circular economy, regenerative economy, and regenerative agriculture that we’re seeing now. We thought about building a way of tracking agricultural products, where they go, and capturing every bit of information about the seed, transportation, the footprint, and everything else. If there could be a way to use Earth observation satellite data for that, that would be it.
Olivier: Do you think anyone else has ended up building something similar to that, or is it still something up for grabs?
Andreea: As far as I know, I don’t think so. It’s a very narrow project idea. It’s more of an ecosystem problem and a systemic topic. So I wouldn’t call it a problem, but it’s more systemic thinking. It’s a new paradigm of thinking that we need to have.
Olivier: You mentioned something pretty interesting as well, that you can be a student but also work at the same time. Hackathons are a way to see what problems we can tackle and what companies are looking for. That’s important to bridge the gap between studying and working. Maybe we could talk more about that—how do you prepare yourself for a career, especially when you don’t know what you want your career to be?
Andreea: Yeah, that’s a really good question. I could talk for at least eight hours about how we’re doing education in the world and how little it prepares the current generation for the real world, for the industry in general. There are several reasons for that. Number one, technology advances so fast that educational systems and curriculums cannot keep up with it. And students—learners in general, and it’s not just students, not just the younger generation; it’s also adults, people who are working already in the field—whose jobs are already being impacted sooner or later by technology because some tasks will be automated, or new types of jobs will be created. They have to get ready for that. Hackathons are one of the things you could do to get out there, see what’s happening. But for me, hackathons are really to innovate for the purpose of innovation. If you want to build a startup or have a business idea, go to a hackathon. But also, if you’re super experienced, and you have skills you want to use, and maybe you’re bored, you can go to a hackathon. Some people do that. But yeah, hackathons are one of the methods for the younger generation to build on their entrepreneurship, but I wouldn’t say it’s the only thing.
Olivier: That reminds me of TechLabs Rotterdam, could you tell me more about that?
Andreea: Yes of course. So just some background information—Tech Labs is a non-profit learning accelerator that started, I think, in 2017 or 2018 in Germany at the University of Münster, and in Rotterdam, it started in 2021. I was approached to start this chapter there. We’re really happy that we’ve run three programs so far—Digital Shaper programs. This Digital Shaper program is actually free of charge to learners who are part of it. We have partners on board who sponsor us to cover the costs. For the learners, it’s really important to be part of a learning community because it’s not just about technology skills. There are a bunch of free courses you can already go online and do yourself. But there’s some statistic that shows that Coursera and all these online MOOCs have the lowest completion rate for their courses online, no matter that they are free—something like 10% to 20% completion rate. So having a learning community—joining a learning community—gives you a bit of accountability, but it also shapes you for the future in terms of communication, teamwork, and exposure as part of the brand of the learning community.
Olivier: Nice! But how do you actually start getting jobs? How do you approach jobs when they say, “You need to have five years of experience,” or “You need to have this and that,” but as a student, I don’t have that much experience. How do I take those first steps to approach that?
Andreea: Yeah, this is a really good question. We have a lot of learners at Tech Labs who are also looking for jobs, or they want to switch jobs, or they want to switch careers from one domain to another. What I’ve noticed is that job descriptions often require a tech stack, cloud technology, a programming language, a methodology like Scrum or Agile, familiarity with other tools for process management, and sometimes five years or more of experience. What I usually advise is: do apply. You don’t have to match all the requirements in the job description as long as you can sell yourself in the motivation letter. Always show a learning attitude and leverage that by saying, “I’m willing to” or “I do score for this. I have some experience here; I’m familiar with this, but I am less familiar with the other requirements, XYZ. However, I am more than willing to learn on the job, and I’m very motivated to prove that I can be a great asset to your team.” That’s something I always advise to people—just keep applying because you never know how your chances are.
Olivier: Amazing advice, that wraps up this episode of the userexperience.org podcast, thanks for coming on the show!
Andreea: Thanks for having me on Olivier.
Congrats on making it to the end of this long post. I hope you enjoyed reading it, I’d love to hear what you think in the comments down below. If you are interested in coming on as a guest on the podcast, send an email to podcast (at) this website’s domain name.
<3 Olivier
Leave a Reply