Supervision

I supervise MSc and BSc thesis projects at TU Delft in constraint programming, search methods, and combinatorial optimization, often aiming for publishable research. On this page, I explain how I work best and what you can expect from me as a supervisor; if you are a prospective student deciding whether you want to work with me, I hope you will find this page informative. Want to get a sense of my previous projects? Then scroll down for a list of projects I supervise or supervised.

Note: I am not a responsible professor. I mentor day-to-day, but official decisions, including your grade or your stay in the program, will require a senior faculty member. What follows is a written description of how I work best as a supervisor, rather than a formal set of rules. I have found that being transparent about my style helps students know what to expect and decide whether it suits their way of working.

❓ Want to address a more specific concern? Please check with supervision FAQ.

Goals

By the end of a project with me, I want you to have:

  • An authentic research experience, rather than merely doing a coursework.
  • Ownership of your topic and process.
  • A path to an academic publication, if feasible given your project output.

My supervision style

πŸ“… Weekly meetings with structured agendas; I am hands-off in terms of format, but please let me know the night before what we will discuss this week.
πŸ—£οΈ Verbal-first feedback: I deliver most of my comments β€œlive” during a meeting, and we work out together the best way to proceed. Of course, for critical deliverables, such as your thesis draft, I will give written feedback.
πŸ—’ Constructive, specific, and high-level feedback: you will receive many critical comments, but I will ground them in concrete, small-scale examples. I also focus my feedback on high-level concerns (e.g., whether your approach is feasible), rather than on minor, technical details.
♻️ Iterative cycle: write β†’ implement β†’ reflect β†’ improve.
πŸ€— Hands-on support without micromanagement: I will not write code or text for you, but you may rest assured that I will help out with either if you need it.

What I expect from you

πŸ’¬ Consistent communication, especially when stuck: remember that I can only help you if you indicate that you need help.
πŸ€” Curiosity, independence, and collaboration: do not expect me to provide step-by-step instructions – I value independence and critical thinking.
πŸͺž Willingness to reflect and adjust: any progress requires changing your behavior; if you engage openly, I’ll spend as much time as needed to help you improve.
πŸ§‘β€πŸ”¬ Active participation in writing and experimenting: I will happily help you set up, execute, and describe a convincing experiment, but if you want to learn it, you will have to do it.

My commitments to you

The points above describe my supervision style in broad strokes. If you are still feeling unsure about how this works day-to-day, check out the supervision FAQ.

  • I take your growth seriously and will support you technically and strategically.
  • I will not disappear on you, block your ideas arbitrarily, or shift the goalposts.
  • I’ll be honest, constructive, and respectful – and I expect the same.
  • If we disagree, I will discuss it openly and work towards a fair solution.

Project portfolio

I have been involved in several thesis projects as a daily supervisor.

MSc theses

Decision Diagram Focused Learning: Efficient Predict-and-Optimize With Decision Diagrams

Jop Schaap, graduated in 2025

Co-supervised with Koos van der Linden

An issue present in many state-of-the-art approaches to predict-and-optimize is the need to re-solve the same problem with different objective coefficients for each gradient computation. In this project, we investigate strategies for predict-and-optimize based on decision diagrams as a representation of a feasible set; since changing the objective values corresponds to re-weighting the diagram, our hypothesis is that we can derive a faster learning routine based on this idea.

predict-and-optimize
decision diagrams
Finding structure in P+O loss functions

Thomas Kuiper, started in 2024

A common trait of loss functions in the predict-and-optimize field is that they give up some information about the problem to make the loss function continuous. To get a better idea of where the strengths and weaknesses lie of P+O losses, we want to find how much of this information on problem structure is retained. In this project, we would like to address the following question: How much of the problem structure do P+O loss functions capture?

predict-and-optimize

BSc theses

Combinatorial Optimisation for Scheduling (2023)

In this project, we have investigated the strategies for guiding a SAT solver into better solution for several variations of resource-constrained project-scheduling problem (RCPSP). A common theme between the subprojects was the importance of the seed values of the VSIDS branching heuristics; despite the common approval of VSIDS as a branching strategy, it has no built-in solution for the cold-start issue. The results of this project suggest several strategies for initializing VSIDS values in a way that triggers a solver into common heuristic solutions.

scheduling
RCPSP
SAT
VSIDS