Supported by a BAEF Fellowship, I obtained a master's degree in Computer Science from Columbia University. In just 9 months, I completed the 1.5-year program, specializing in advanced Machine Learning topics.

At the Creative Machines Lab, I worked on the exact parameter reconstruction of black-box neural networks. By visualizing the local optima in the reconstruction process, I came up with an advanced sampling strategy iteratively combining in-distribution and out-of-distribution data. My sample-efficient approach stabilized the reconstruction process for large networks and enabled me to reconstruct a network with 1.5 million parameters, the largest reconstruction seen to date. The corresponding paper is currently under review for publication and the preprint is available on arxiv.

Before that, I earned a master's degree in Computer Science Engineering from Ghent University, where I acquired competitive software development skills and developed a thorough theoretical foundation in Machine Learning.

During my thesis research, I worked on domain adaptation of Computer Vision systems to out-of-distribution data. I applied feature embeddings produced by multimodal transformer foundation models to cluster out-of-distribution data and automate the labeling of newly acquired data. My efficient framework labeled 10,000 data samples in under 2 minutes and achieved accuracies of over 99% on both state-of-the-art datasets and production datasets provided by Robovision, making an automated retraining system feasible in the industry for the first time.