I just counted: 17 students supervised over the past few years. Master’s theses, doctoral co-supervision, internships. Topics ranging from compiler design to cirrhosis prediction to brain-computer interfaces.
Every single one taught me something I didn’t expect.
The computer science student who changed my research direction:
Ward Ceyssens came to me wanting to work on cross-subject BCI decoding. I thought: “This is nearly impossible. Neural signals vary so much between people, we might not get anywhere.”
He didn’t care. He dug into transfer learning approaches, domain adaptation techniques, ways to find generalizable patterns in ECoG data. His work is now under review at IEEE Transactions on Biomedical Engineering, and it’s opened up an entirely new research direction for federated BCI systems.
Lesson: Sometimes the “impossible” project is exactly the right one.
The statistics student who taught me about clinical thinking:
Meseret Assefa Kerga’s thesis on cirrhosis survival prediction forced me to think like a clinician, not just a machine learning researcher. She kept asking: “But what would a doctor do with this prediction?”
Her insistence on clinical interpretability—not just AUC scores—made the work genuinely useful instead of just technically impressive. That shift in perspective now influences all my healthcare AI projects.
Lesson: Students from different backgrounds see blind spots you didn’t know you had.
The compiler enthusiasts who kept my passion project alive:
Denzell Mgbokwere (optimization and type checking) and Robert Rysskin (bytecode redesign) are working on Cardinal, my C++20 Wren VM reimplementation. This project has nothing to do with my main research, but it’s my way of staying connected to programming language theory—my first academic love.
Their enthusiasm reminded me why I fell in love with computer science in the first place. Some weeks, working with them on type systems and bytecode is what keeps me energized for the healthcare AI grind.
Lesson: Let students help you stay connected to what you love.
The doctoral student I co-supervise but don’t fully understand:
Anh Phuong Do’s work on individual reference intervals for clinical event prediction is deeply statistical. Like, really statistical. Bayesian modeling, longitudinal data analysis, stuff that makes my CS brain hurt.
Co-supervising her (with Prof. Liesbet Peeters) has been humbling. I contribute the machine learning perspective, but I’m constantly learning from her statistical rigor. Our paper is under review at IEEE JBHI, and honestly, she’s taught me as much as I’ve taught her.
Lesson: Sometimes your job is to learn alongside the student, not always be the expert.
The patterns I’ve noticed:
Cross-subject BCI (Ward) Federated tensor regression (my own postdoc work, inspired by student questions) Using BTTR for time series imputation (Mohsen)
Biomedical students: “Would a doctor actually use this?” Statistics students: “Are we accounting for confounding properly?” CS students: “Can this scale to real-world data volumes?”
I’m still learning to recognize which is which earlier in the process My default is to give space (because that’s what I needed as a PhD student) But some students thrive with weekly check-ins and concrete milestones
Students dealing with impostor syndrome Students struggling with data that won’t cooperate Students facing personal challenges while trying to finish Sometimes my job is less “technical guidance” and more “you’re going to be okay” What I’m still figuring out:
How many students is too many? I’m supervising 6 master’s theses this year while coordinating research projects, teaching 4 courses, and doing my own research. Some days I worry I’m spread too thin. Am I giving each student enough attention?
When to say no to a topic I don’t fully understand? I’ve supervised students working on areas adjacent to my expertise (household wealth estimation from remote sensing, transformer models for time series). How far can I stretch before I’m doing them a disservice?
How to balance support with appropriate challenge? I don’t want to make things too easy. Struggle is how you learn. But I also don’t want to leave students drowning. Where’s the line?
The privilege:
Despite the challenges, supervising students is the most rewarding part of my job. More than papers, more than grants, more than conference talks.
Watching Qiang Sun (exoskeleton BCI) and Eva Calvo Merino (finger movement decoding) publish from their doctoral work—that matters. Seeing former internship students (Iris, Didier, Aurelie) go on to their own research careers—that’s the impact.
Every student brings their own perspective, their own questions, their own way of seeing the problem. And every single one has made my research better.
To my current and former students: Thank you for letting me be part of your journey. You’ve shaped mine more than you know.
To fellow supervisors: What have your students taught you?
Currently supervising: 6 master’s students (CS, Statistics, Data Science) + 1 doctoral co-supervision. Topics: federated learning, compiler optimization, cardiovascular health, time series, arrhythmia detection. Always learning.
#StudentSupervision #AcademicMentorship #ResearchEducation #PhDLife #ThesisSupervision