AI & Data Science graduate · Medical Imaging · Computer Vision · Machine Learning · Nijmegen, the Netherlands

I'm an AI engineer at heart with a background in deep learning, medical imaging and language technology. I care most about AI that puts people first: systems that help clinicians, support learners, and help us understand ourselves a little better.

My interest in AI started with a question about people: how do we make sense of the world, and can a machine learn to do the same? That question took me through a bachelor's and master's in Artificial Intelligence at Radboud University, with courses in cognitive psychology alongside machine learning.

• BSc Artificial Intelligence (Radboud University, 2012–2016) My thesis explored whether a neural network could learn human moral judgements about care and harm from survey data, a bottom-up approach to artificial moral agents. Grade: 8.5.

• MSc Artificial Intelligence, Intelligent Technology (Radboud University, 2017–2021, graduated with an 8.0) For my thesis at the Radboudumc Diagnostic Image Analysis Group I built a deep learning system (U-Net) that detects the abdominal aorta in handheld ultrasound and measures its diameter, running in real time on a smartphone. The goal: letting non-specialists screen for abdominal aortic aneurysms outside the hospital. Grade: 8.5.

• MSc Computer Science, Data Science (Radboud University) Coursework completed. During an internship at Thirona I researched self-supervised learning (SimCLR) on chest X-rays for tuberculosis detection, showing that pre-trained representations can cut the number of labels needed roughly in half.

• Software Engineer at Binding Blocks (2024–2025) Built an educational algebra and calculus game with symbolic equation solving and a step-by-step hint system (bindingblocks.nl). I also built the math learning platform Wisgame.

My preference has always been medical AI: models that give a clinician a fast, reliable second opinion, or bring diagnostics to places where specialists are scarce. But I love every application where the human is at the centre: education, language, wellbeing, and making AI understandable instead of a black box.

Wisgame is a Dutch math-practice game I built on my own (107 commits, 2022–2025). It generates algebra and calculus exercises, checks typed answers and shows a worked solution step by step in Dutch. An independent code review (October 2026) concluded:

The project shows solid self-taught engineering ability: a working symbolic-math engine designed and built from scratch, without a computer-algebra library, and shipped as a live web app. Overall level: junior-to-medior engineer with strong algorithmic and CS fundamentals, whose main growth areas are software-engineering practice (tests, structure, security) rather than problem-solving.

Technical highlights

• Symbolic math engine from first principles (~1,200 lines): expression-tree classes with exact rational arithmetic and operator overloading.
• Data-driven rewrite system: simplification and differentiation rules live in JSON, match tree shapes by path, support numeric side-conditions, and need no code change to extend.
• Explainable output: every rewrite step carries a Dutch explanation ("Beweeg 5 naar rechts", "Deel kanten door −2"), so the tool teaches rather than only grades.
• Robust answer checking: numerical comparison for derivatives, and Hausdorff distance between solution sets for equations, so x = 3 ∨ x = −3 matches x = −3 ∨ x = 3 (distance function written with help from ChatGPT, then adapted).
• LaTeX parser (recursive descent, built with help from ChatGPT) that turns student input into the engine's expression trees.
• Full web stack: Flask, Flask-SocketIO, SQLite, MathJax/MathLive, matplotlib-generated SVG diagrams.
• Continuous deployment: GitHub Actions deploys every push over SSH to a DigitalOcean server running a systemd service.

Tested: the differentiation engine solved 14 of 17 level-1 exercises correctly, including the quotient rule (−2/(x−1)² in 13 steps); the equation solver handled linear, factorable quadratic and square-root equations.

Review written with Claude (Anthropic's AI assistant) after reading the source and running the engine. Not an official certificate.

Machine learning: deep learning, segmentation, classification, self-supervised learning, interpretability Computer vision: medical imaging, ultrasound, X-ray analysis Language: NLP, information retrieval, speech recognition Programming: Python, Java, MATLAB, JavaScript, PHP, SQL Tools: PyTorch, TensorFlow, NumPy, Pandas, Git, Linux

I'm working towards a role as an AI Engineer, Data Engineer or Data Scientist.

Right now I'm looking for a training programme, a learning-work placement or volunteer work where I can put my skills to use, keep growing, and contribute to something meaningful. If you're working on human-centred AI, especially in healthcare or education, I'd love to hear from you.

• ✉️ dljvanerp@gmail.com
• 💼 LinkedIn
• 🐙 GitHub