Machine learning & data engineer · MSc Computer Science, TU Delft · Eindhoven
From pollen to people to postings: I find structure in messy real-world data, and ship it.
Open to ML, data science and software engineering roles in the Netherlands
Classified nettle-family pollen from whole microscope Z-stacks rather than flattened images, building 3D convolutional networks in PyTorch.
95% test accuracy with a ResNet3D on full stacksFused wearable positioning and audio from an inclusive preschool, discovered activity contexts with unsupervised learning, and compared children with and without hearing loss using mixed-effects models.
Inclusion effects invisible in the pooled analysisA public job board that reads postings from Dutch employer career sites, classifies tech roles and extracts skills, seniority and visa sponsorship. Designed, deployed and run in production on Kubernetes on AWS, with an evaluation gate in CI.
techjobsradar.nlDifferent data each time; the same instinct to turn noise into something people can use.
Open-source research applying synaptic tagging and capture, a theory of how memories become permanent, to neural networks, with a consolidation log that explains what the network retains.
A TU Delft group study of whether three LLMs labelling meeting transcripts agree more when they see each other's earlier labels; agreement rose from 0.185 to 0.216.
Matches messy bank-statement text to merchants and detects recurring payments, reaching a B-cubed F1 of 0.95 on held-out data.
A loss-free data-ingestion proof of concept for a smart-home system, built with FastAPI and RabbitMQ and tested under simulated network outages.
Reimplemented an underwater image-enhancement network in PyTorch from its paper and reported which results did and did not replicate.
Real-time hand-gesture effects in the browser with MediaPipe, and a brushable D3 timeline in a linked earthquake and tsunami visualisation.
Native English, conversational Dutch. Off the keyboard: basketball.