Research
Reproducible, interpretable, open source.
I build AI-driven neuroimaging methods for stereotactic and clinical use, where millimetric accuracy constrains model design, uncertainty is explicit, and automated workflows remain auditable and reproducible across scanners, sites, and populations.
Lane 1: Measurement & QC
Make errors visible, comparable, and actionable.
Lane 2: Targeting & Anatomy
Interpretability and uncertainty over “pretty maps.”
Lane 3: Reproducible Tooling
Pipelines that survive other people’s hands.