He came to build an app and became a researcher

“I never saw myself as an academic or a researcher. I always wanted to create my own tech startup.”

So says Phongsakon Mark Konrad, known simply as Mark. He arrived in Sønderborg in 2022, aged 25, with ambitions to build his own business.

He had left a mathematics and economics degree in Germany to develop an AI-assisted fitness app. When SDU launched an English-taught bachelor’s programme in Software Engineering in Sønderborg, he saw an opportunity to learn how to build the software his business would depend on.

During his studies, however, he developed a taste for research. Today, he wants to continue in academia and eventually lead his own research group.

He is particularly interested in causality in artificial intelligence, that is, whether AI can tell the difference between things that merely happen together and things that actually cause each other. Modern AI models are good at finding patterns in vast amounts of data, and they can often answer questions about cause and effect correctly. But it is unclear whether they truly understand what causes what.

“The way we train models today is mostly statistics and correlation. The models are very capable, but they are also enormous and require a lot of resources. It works for now, but it is not elegant,” he says.

Mark wants to help solve these fundamental problems. His interest grew during his degree in Sønderborg, where he joined a research group after just one year of study.

An email in his first semester

During his first semester, Mark wrote to Serkan Ayvaz, Associate Professor at SDU’s Centre for Industrial Software, part of the Mærsk Mc-Kinney Møller Institute. Ayvaz had introduced his work in artificial intelligence, and Mark had begun to consider a research career and a PhD. He asked for a list of books and courses he should work through. He received the list and got started.

“I think he realised that I really meant it,” says Mark.

At the end of his second semester, Ayvaz offered him a position as a research assistant. Such positions usually go to master’s students, but the Software Engineering programme in Sønderborg was new, and Mark belonged to its first cohort.

“A lot comes down to luck and timing. There were no master’s students, so he had me,” says Mark.

His task was to predict road conditions in agricultural areas using computer vision, a field he had never studied. He had to learn it from scratch.

“It was very hard at first. But he explained the problem and suggested how I might approach it. After that, I was free to explore. He trusted me, even though he knew I did not really know anything yet.”

Research felt familiar

The work turned out to have much in common with what had attracted Mark to entrepreneurship.

“What struck me most about research is that it resembles entrepreneurship. You have an idea and a problem you want to solve. You build a proof of concept, start small and improve your idea step by step,” he says.

He brought that approach with him from the programme’s semester projects. Students work in groups of six on assignments from companies such as Danfoss and LINAK. They use agile methods such as Scrum, with weekly meetings and documentation, just as they would in industry. Mark worked with the same group from his second semester onwards.

“It was painful at times, but I had a really good group. You learn project management and communication, and we presented our projects over and over again. In the first semester, we were very nervous. By the end, we were completely calm,” he says.

He believes this practical experience distinguishes the programme from many others. A friend in Germany studied Scrum in theory and was assessed through multiple-choice questions. In Sønderborg, the students had to put it into practice.

“People have this image of software engineers sitting alone behind a screen and never talking to anyone. But that is not how it works. To be a good software engineer today, you need to be able to code, but you also need to communicate. You always work in a team, and you need to understand the client’s requirements and explain what is possible and why.”

Ten papers and the Super Bowl of machine learning

Mark made research a priority throughout his degree. The result is ten peer-reviewed research papers, seven of them with him as first author. He has presented his work at international conferences, including workshops at the machine learning conference ICML in South Korea and the KES conference in Japan, and other researchers have begun citing his papers.

Recently, one of his papers was accepted in the main track at NeurIPS, one of the world’s most prestigious machine learning conferences.

“That was a dream I had set myself for 2030. NeurIPS is pretty much the Super Bowl of machine learning research,” he says.

The bachelor’s project he completed with fellow student Tim Lukas Adam was also named the best among a cohort of more than 100 students. Along the way, he found time for an exchange at the Hong Kong University of Science and Technology.

When AI becomes better at coding than you are

The rapid developments in artificial intelligence have also prompted Mark to consider where he can make the greatest contribution as a researcher.

“I love and hate AI at the same time. I used to enjoy coding, but AI has become so good at it that it is better than me. Coding everything from scratch for my first paper took a year. Today, it could be done in a week,” he says.

With AI helping to write code and run experiments, Mark wants to spend more time on the theory behind the models. That requires a stronger mathematical foundation.

“Even with all the AI tools, you need to be able to judge whether a model’s output is correct. That takes years of training. And if I were starting again, I would probably have studied mathematics or physics,” he says.

He adds that he may have read too many biographies of his idols, such as Richard Feynman, Einstein and Oppenheimer.

“They were all theorists. Applied research is exciting and has an impact, but it always comes from theory. If you want to solve fundamental problems, that is where you need to go.”

Cambridge or Tübingen

This summer, Mark was offered places at the University of Cambridge (MPhil in Machine Learning and Machine Intelligence), the University of Tübingen (MSc in Machine Learning) and the University of Copenhagen (MSc in IT and Cognition). Cambridge had been his dream, but conversations with people from the university at a conference made him question whether its 11-month programme would give him enough time to explore the subject in depth.

“I have nothing against stress, but I want to become a really good researcher, and that takes time,” he says.

He chose Tübingen, one of Europe’s leading centres for machine learning, with Max Planck institutes in the city and the Cyber Valley research network. Its two-year, research-oriented master’s programme allows Mark to spend a year and a half strengthening his mathematical foundation.

He has also been awarded a Research-Oriented Master’s Scholarship from the Zuse School ELIZA, a German graduate school in AI that is part of ELLIS, the European Laboratory for Learning and Intelligent Systems, a pan-European network of machine learning researchers. The scholarship prepares students for doctoral research in machine learning and lets Mark focus on his studies and research.

“It is the best decision I have made so far. The master’s is just a step along the way. I am thinking long term,” he says.

Easy access to researchers

Looking back on Sønderborg, Mark particularly values a study environment where lecturers and researchers are easy to approach.

“Everyone here replies and takes the time. It is easy to get in touch with professors and have a normal conversation with them. I do not think you get that on a large campus with 20,000 or 40,000 students,” he says.

As a member of the programme’s study board, he has also seen students’ concerns taken seriously.

“If you explain the reasons for your criticism, they listen. The programme has developed very quickly, and you can have a lot of influence if you take your role as a student seriously.”

The research group around Ayvaz has grown from Mark alone to around ten people, and Mark will continue collaborating with the group from Tübingen. He warmly recommends the programme to students who want practical experience in software engineering.

“If you are a software engineer at heart and enjoy practical work, it is a really good programme. Serkan spent ten years in industry in the US and understands the challenges. If you go too far into theory, you can sometimes lose sight of reality,” he says.

For now, Mark has put his startup plans aside. He sees himself spending the coming years in academia and hopes to become a professor with his own research group one day.

“I never saw myself as a researcher. That is something I discovered here.”

“I would also recommend Sønderborg to students who want to pursue a research career, especially in industrial and applied research. The labs focus on challenges from industry, and even as a first-year bachelor’s student, you can get in touch with researchers, explore your interests and contribute to research,” says Mark.