Hello, I'm Tymo van Rijn
Deep learning enthusiast focused on medical AI — I build systems that help make sense of what static images miss.
During my graduation internship at the Idiap Research Institute in Switzerland, I built a temporal deep learning system for automated grading of retinal inflammation in fluorescein angiography. My thesis, "Beyond the Last Frame: Temporal Deep Learning for Automated Grading of Retinal Inflammation in Fluorescein Angiography," combined a Vision Transformer with a GRU sequence model to read an entire angiography sequence instead of relying on a single frame — and I graduated with a 9.6.
Deep learning is what pulls me in, and medical AI specifically is where my passion lies. There's something uniquely motivating about building models that can help clinicians catch what's easy to miss — a subtle pattern in a scan, or a change that only becomes visible across time rather than in a single image.
Outside of research, I'm the proud owner of Prozum Labs (prozum.nl), where I built BaseLine — a tool that analyzes your lab results against your own personal baseline instead of population averages, using the reference change value method labs have relied on for decades.
None of this ships without solid software engineering underneath it, which is where my web development background comes in. Next.js, TypeScript, Python and Flask are the tools I reach for to turn a model or an idea into something people can actually use — this blog included.
My Mission
My goal is to build AI systems that support clinicians and improve patient outcomes — starting with the automated retinal-inflammation grading work from my graduation thesis at the Idiap Research Institute, and continuing through what I build next.
I believe in rigorous research, sharing what I learn, and shipping software that actually gets used. Prozum Labs and BaseLine are a direct expression of that: giving people a clearer, more personal way to understand their own health data.
Through this blog, I aim to document that journey — the research, the code, and everything in between — for anyone on a similar path toward deep learning and medical AI.
"The best way to predict the future is to create it."
— Peter Drucker