About

Who

Hey, I am Paul! I currently work at KLM Dutch Royal Airlines, developing C++ tooling. My interests are mainly in C++, low-latency code, algorithms, but also computational models and machine learning.

During my master’s degree, I conducted research in compression and pattern recognition at the Institute of Physics at the University of Amsterdam. I was also involved in geometric deep learning research with the Amsterdam Machine Learning Lab on point clouds, focusing on reducing the desciption length of the representation. Before that, I taught other graduate students computational modelling and worked in digital risk at Deloitte.

I am motivated by learning new things, especially if they seem much too complex initially. In my head, I think about my personal influence on my success as a Lévy process. I can only influence the work I put in and additional noise factors influence my success probability. I cannot force success on a particular event, but I can influence the probability of success by working on myself - and there is always more to learn. Success on events is a Poisson-process: I can only increase the average occurrence rate of success given an event. A bit like this,

where

is the drift caused by working on myself.
is the noise caused by factors outside of my control.
compound Poisson jumps, moddeling rare but impactful events.

Of course, I my efforts are not fully deterministic and we could add a lot more coefficients (volatility, frequency of success-influencing events etc). However, this is a model that is useful to me - which is what modelling is all about: usefulness. It keeps me motivated and abstracts the success from the effort, making me optimise for a high drift coefficient, rather than an outcome outside my control. Which is probabily the most techical description of a growth-mindset I could have given.

Why

I started this blog as a repository of knowledge, mainly for myself. I enjoyed teaching and I love learning. Depending on the topic and my mood, I may choose to write a more technical paper or an approachable blog. If you find something useful - great! If you find something incorrect, even better. Send me a message on LinkedIn to contribute to my drift coefficient as well.