Finn S. Fassbender

I’ve spent the last few years doing a double degree in Medicine and Machine Learning—mostly because I couldn’t pick one, and the parallel study program let me get away with it. It turned out to be a good fit.

Spending time in both the hospital and the lecture hall, I realized the questions I care about most live in the gap between what we can predict and what we can actually change. In medicine, you’re constantly making decisions under uncertainty, and the stakes are high.

That’s what pulled me into causal inference. Not because I wanted to build slightly better predictors, but because I want to understand whether an intervention actually helps this patient, not just the average one. A model telling me it’s “uncertain” isn’t that helpful unless I know why. Is the patient unusual? Is the data messy? Without knowing that, I’m not taking a calculated risk. I’m just guessing about a guess.

With personalised medicine, the goal isn’t to ignore the evidence from thousands of others and just back a hunch. It’s to build a model representation that helps me see how the person in front of me is similar to, or different from, the populations in the trials I’ve read. Electronic health records are essentially a huge log of interventions and outcomes. We just haven’t been good at reading them causally. Yet.

When I’m not in the lab or clinic, I spend time on projects that probably won’t improve my h-index but might make medicine a little less broken. I co-lead Ankizin, Germany’s largest free, community-owned medical flashcard deck. It started as frustration with paywalled study materials and is now used by over 22,000 students.
I’m also the National Officer for Research for the bvmd (the German Medical Students’ Association), where I’m working on improving doctoral training structures. Right now, the system expects med students to do PhD-level research without clear structure or funding. It’s something we must improve.

Medical education can feel slow to change. We often train students to memorize facts rather than navigate uncertainty. I’m interested in tools that help shift that balance.

At the end of the day, I just want to help build a kind of medicine that’s a bit more precise, a bit more equitable, and a little more curious. And I really do want those uncertainty estimates I can actually use.