After there were already master’s tracks that embraced AI in various separate TU/e programs, seven TU/e departments joined forces to start the new master’s in Artificial Intelligence & Engineering Systems (AI&ES) together in 2021. In the meantime, the first engineers have graduated and received their diplomas. Thanks to the buzzword AI in their program, the interest in these engineers is enormous. And after talking to three of them about their expectations beforehand, their experiences with the master, and their dreams, that demand can only get bigger.
Another new master’s program and one that is run by no less than seven departments (Electrical Engineering, Mechanical Engineering, Applied Physics, Mathematics & Computer Science, Biomedical Engineering, Industrial Engineering & Innovation Systems, and Built Environment). What did they expect from their master as a student, and what has it brought them’
Choosing AI&ES
Together with brand new masters of science, Sem Draaijer, Alexander de Pauw, and Teun van de Laar, we went back in time and asked why they chose this brand new master at the time. Sem Draaijer: "After my bachelor’s degree in mechanical engineering at the Amsterdam University of Applied Sciences, I wanted to continue my studies.""I was also looking for a little more depth in my studies. I found it at TU/e, where there was an AI master’s program in Mechanical Engineering. After I completed my pre-master, it was merged into this new master."
After a bachelor’s degree in Applied Physics, where Teun van de Laar was also a student assistant in the course ’Machine Learning’, he immediately opted for the master’s degree in Nuclear Fusion. "No matter how interesting that field is, innovation and especially its application in practice is still very far away."
"So when I saw that the AI&ES master was starting, I decided to make the switch. This master’s was exactly in line with what I already liked about the course ’Machine Learning’ at the time. And, at the same time, this master was broader than the already existing master in ’Data Science’ thanks to all those different faculties. And I haven’t regretted it for a second," Van de Laar says.
Even during my previous studies, I followed the news about AI closely. Especially the applicability in all kinds of areas, from healthcare to industry, appealed to me. Alexander de Pauw
After high school in Italy, Alexander de Pauw obtained a bachelor’s degree in Liberal Arts and Sciences in Maastricht. "That’s also where I discovered my love for AI. Even during that training, I followed the news about AI closely. Especially the applicability in all kinds of areas, from healthcare to industry, appealed to me," he says.
"After that, I was looking for a master’s degree in AI, but preferably broadly structured and focused on applying AI models, and I found that here in Eindhoven. In addition, the campus really appealed to me, so that’s how I made my choice," De Pauw concludes.
Estimating Cardiac Output From Arterial Blood Pressure Using Deep Learning
More than a year into his studies, Sem Draaijer chose the Healthcare track because he is intrinsically motivated to do something for healthcare with his acquired knowledge of data and information systems. He graduated in the Computational Biology group of Natal van Riel within the Department of Biomedical Engineering.The focus of his thesis was on measuring and predicting the output of a heart as well as how to predict the risk of stroke based on arterial blood pressure.
Draaijer: "In the Intensive Care Unit (ICU), it’s essential to monitor patients’ Cardiac Output (CO), which is a measure of the amount of blood the heart pumps per minute. Proper control over this value can help improve treatment decisions and patient outcomes."
"Previous research has shown that a specific type of neural network (1-D CNN) can be used to estimate stroke volume (SV). This is the amount of blood the heart pumps per heartbeat, calculated from the blood pressure wave and information about the patient"
"In our research, we further improved this model by adding extra steps, such as better data filtering and new features from the data, and by investigating various advanced models. The combination of new advanced models and features from the blood pressure waves has resulted in an improved performance compared to the existing model. This research shows that further development of deep learning algorithms can help to estimate CO more accurately with blood pressure waves."
Learning about AI in engineering practice
Draaijer: "I hadn’t worked with AI before in my studies. It was very nice that each course always started with the basics, making it easy for everyone from different backgrounds to pick up. Quite early in the program, we chose our graduation track. I chose the Healthcare track because I find the combination of people and technology exciting."Compared to Draaijer, Van de Laar had a more leisurely start in theory because he was already a student at TU/e. Van de Laar: "I knew my way around the TU/e. What struck me most about this master’s was the many practical cases and projects it offered. As a matter of fact, every subject offered practical examples. It was special and super instructive to see how to use the material you learned in practice."
I found the broad basis that we are given very valuable, from data science to ethics. Teun van de Laar
"I already had a foundation in neural networks and knew the basics of AI. But when you see what you run into when you start using that theory to solve a real case, you start to understand what you encounter in practice. In addition, I found the broad basis that we are given very valuable, from data science to ethics, which is also very important," Van de Laar adds. He chose the Mobility track for his thesis.
De Pauw came to the master’s program with an interest in AI and robotics after he had previously studied people. "I also really enjoyed the master’s, and I liked that the first courses included a practical side. We learned so much there that I still use every day, such as a good foundation with coding and GitHub (support software for software development, ed.)," says De Pauw.
"I also find the university’s connection with industry so valuable. It’s really super easy to go from the university to different companies and talk to them. I chose Hightech Systems and Robotics and ended up graduating on a It’s great that the university facilitated that."
VernaCopter: Natural Language-based Drone Control using Large Language Models and Formal Specifications
His choice for the Mobility track, says Teun van de Laar, was mainly motivated by the breadth of the program and the subjects. For his graduation project, he investigated how drones can be controlled with natural language or the language we humans speak among ourselves.For example, you could ask a drone to ’inspect the chimney’s lead flashing on the left roof’. To do this, he delved into the world of large language models (LLMs, ed.), better known by brand names such as Chat GPT and CoPilot.
"In my report, I introduce VernaCopter, a system that enables non-technical users to control drones (quadrocopters, ed.) using natural language. Signal Temporal Logic (STL) functions as an intermediate representation of tasks specified in natural language. The LLM is responsible for task scheduling, while formal methods handle movement scheduling and address the limitations mentioned earlier. We use automatic checks, based on LLM syntax and semantics, to improve the quality of STL specifications."
"We tested the system’s performance in experiments in different scenarios, different user engagement, and with and without automatic checks. These experiments showed that engaging the user in a conversation improves the drone’s performance. We concluded that the LLM we used made a difference in performance, but the controls we added did not always improve performance due to the many incorrect corrections."
Graduated and then out into the world’
After graduating, Draaijer and Van de Laar wanted to take time to unwind and see where they wanted to get started. Draaijer: "I first started looking very broadly. And there are all kinds of requests coming your way."Van de Laar: "I’ve had a lot of conversations, and it’s a search. There aren’t any vacancies explicitly requesting our master’s background yet. So you always have to explain what your field of expertise is."
"That is very recognizable," says Draaijer. "But I’m done with that now. I now know that I want to return to healthcare in the long run, but first, I want to look around some more in retail. That’s why I’ll start as a Models Analyst Forecasting at Coolblue in December."
You don’t have to worry about not being able to get started with this master’s anywhere, but finding a job that matches your ambitions requires a bit more searching. Teun van de Laar
"Indeed, you don’t have to worry about not being able to get started with this master’s anywhere, but finding a job that matches your ambitions is a bit more searching. I’m also looking more specifically now, and I’m now talking to the right parties, I think," Van de Laar adds.
Both will be able to get started soon, something De Pauw is not worried about. He started his own company Texterous with his friend and graduation partner during his studies.
"We have benefited a lot from the TU/e’s honors program and the connections we have been able to make with other entrepreneurs through teachers, the TU/e Contest, and EAISI. That’s how we even found some of our current customers," says De Pauw.
"With Texterous, we help institutions at home and abroad to use AI in their business processes, but in a secure way. That’s going very well. But my dream remains to one day be able to work at the European Space Agency. So if they call, I will stop working at my company."
Enhancing Coffee Production Efficiency Through AI-Based Predictive Maintenance of Grinding Rolls
Thanks to the contacts of a teacher, Alexander de Pauw and his partner were able to do their final research at a company. They ended up at Jacobs Douwe Egberts, where they were allowed to model and optimize the grinding rollers, which make filter coffee from roasted beans.
In addition to professional knowledge, this provided very valuable insights into the role of a data scientist or AI expert in a business situation. This is know-how that also comes in handy for De Pauw in his own company.
De Pauw: "My research aimed to determine the optimal time to replace grinding rollers in a coffee production environment. To do this, we analyzed and predicted the loss of efficiency over time."
"Using data and constraints from a coffee production company, we developed a predictive model in our study. This model can predict when the efficiency of the rollers decreases to a critical level. This lets the model indicate the best time for replacement."
"The goal is to maximize the service life of the burrs while optimizing production speed and electricity consumption. The research has shown that efficiency losses are a predictable factor in this process, and deep learning models can be used effectively to predict these losses. These predictions can inform decisions about both the optimal grinding strategy and the timing of roll replacements."
Content masters
The three graduates are unanimous about their positive assessment of the master’s. Are there still things they missed? De Pauw can’t think of anything, but both Draaijer and Van de Laar would have liked to have interned in a company.I would have liked to have had the experience of what it is like to work in an company’s department. Sem Draaijer
"I would have liked to have had the experience that Alexander (De Pauw, red.) has gained by working in a company’s department," Draaijer concludes.
Van de Laar agrees: "That would have been valuable experience. Especially now that we are applying for our first job. You have a better idea what a role or vacancy entails if you have worked at a company."
"Such an external master research "
