I am a postdoc in AI, working on causal- and physics-aware machine learning with the ADA research group at Leiden University. My general research interests lie in machine learning, and the occasional black-box optimisation, in scenarios where performance metrics and fitness function values cannot be relied on, which causal machine learning and physics-aware machine learning are aimed at. I also have a strong preference for working on problems with a high societal or environmental impact that we cannot already solve ourselves (as humans). As a result, I generally work with spatio-temporal data, usually with an emphasis on urban and environmental applications through Earth observation data.

What to expect on this website

• Major research directions and their motivation. As a postdoc, I am working on causal machine learning for spatio-temporal data, where I focus primarily on Earth Observation data. During my PhD I worked on physics-aware machine learning for Earth Observation data, particularly focusing on physical model inversion.
• Main-author papers and their associated resources, with a plain-language summary. The plain-language summary should be an accessible way to interpret the paper for an intelligent and well-educated, but non-expert, audience (more on this in the motivation below). Otherwise, you can expect a link to the publication, the abstract, a link to the code repository and an “updates” section, in case more relevant developments occurred after the initial publication.
• Guides for how to use specific algorithms I developed. I often find it hard to use a method published in some paper, because there is a lot of overhead getting familiar with the details of the method. I would like to help people who may feel the same way about my own papers by including some instructions and explanations here, with a focus on ease of use for any problem (so beyond just reproducing results on a specific dataset, for which the scripts are already available).
• A good old publication list.
• [Planned] Some Korean-language content. I’m not sure yet how I want to do this, but I would love to somehow contribute to promoting further scientific collaboration between the European Union and the Republic of Korea, which I feel like I’m too unusually well situated for to do nothing at all with.

Why I made this website

The direct cause for deciding to make this website was that I did a little informal experiment with LLMs. I know from experience that interpreting academic papers and using their methods can have a high barrier of entry, requiring a lot of time, effort and frustration. It would be tempting to use some LLM to summarise a paper, and ask questions about it. This is a task that I, despite being generally fairly skeptical about the abilities of LLMs, felt they should be quite good at: changing the text of a concrete text source into a more interpretable format. So I asked several prominent models to explain one of my own papers to me, in the vein of “LLMs sound very convincing until they start talking about something you know well”.

I was shocked by how completely wrong it was. I don’t mean there were some small mistakes, or significant artefacts/hallucinations on key parts of the paper. I mean that it was actually completely wrong about basically everything it said. The more questions I asked, the worse it became. Under the assumption that this is an interface through which many readers may interact with my research output, I felt that it would be better to create a website where I gather all the resources in one place, and I take special care to explain the core concepts and findings of my papers in a format anyone can understand.

As I grew as an early-career scientist over the years, and was influenced in part by colleagues with a particular enthusiasm for science communication and public outreach, I learned that producing research output is not the only task for scientists. We also have an obligation, as publicly funded researchers, to disseminate this work back to the general public. Not just in the form of open-access research papers, that really only experts can interpret, but in a form that is accessible and understandable to anyone.

I won’t claim to have achieved this goal, or even that I did a good job yet through this website. But I have to start somewhere, and this is the first step I will make in this direction.