Clinical ML and AI/ML products that ship — and stand up to scrutiny
I split my work between careful Clinical ML research and shipping AI/ML products. The work I care about: calibrated risk, decision-curve thresholds, and figures that anyone on the team can read. I would rather ship fewer claims with better evidence.
Calibrated risk •Decision-curve thresholds •External validation •Reliability diagrams •Conformal options •Model cards •Dataset cards •Reproducible figures •Accessible UX •Lighthouse-tested on every PR •Prospective pilots •Calibrated risk •Decision-curve thresholds •External validation •Reliability diagrams •Conformal options •Model cards •Dataset cards •Reproducible figures •Accessible UX •Lighthouse-tested on every PR •Prospective pilots •
Tested per PR
Lighthouse
Weekly
Ship cadence
AA target
Accessibility
Green
Tests
Where I can help
AI/ML products
- Production-ready web apps with accessible, clean UX.
- Observable data pipelines and evaluation harnesses.
- A/B or counterfactual checks to prove real lift.
Explore
Clinical ML research
- Calibrated risk with decision-curve thresholds.
- External validation, reliability plots, uncertainty.
- Model/Data cards and reproducible analysis.
Explore
Selected work
Bench → bytes
Research
Bench assays paired with tidy analysis pipelines and clear figures.
Open
Open & reproduciblePublications
Preprints and method notes with code, data, and repeatable results.
Open
MeasuredPerformance
Budgets, Lighthouse, and live device vitals — fast on purpose.
Open
Let’s buildCollaborations
Scope, guardrails, and weekly milestones with steady progress.
Open
Have an idea that fits?
Send the smallest useful dataset and a short note. I’ll reply quickly with next steps.