My team is building an AI pipeline that finds choroidal nevi in the fundus images clinics already take during routine eye exams.

These nevi are easy to miss, which is why they so often are. The ones that slip through can be the dangerous ones, the few that turn cancerous. Finding them reliably is hard, so we're building a different kind of model to do it.

I don't write the algorithm myself; my job is to keep a team of clinicians and engineers focused on one question: would a real clinic actually use this? That question ruled a lot out. We couldn't ask clinics to buy new hardware, key in more data, or add clicks to an already busy day. It's taken time, trust, and a lot of iteration to get here, and we're close to launch.

I'm a health services researcher at the University of Calgary, where I co-lead O3R, an ocular-oncology outcomes-research lab. The work above is peer-reviewed; you can read it.