Collaborations that ship — and stand up to scrutiny

I work on two tracks: AI/ML products (clean UX, measurable lift, real users) and Clinical ML research (calibrated risk, honest figures, transparent thresholds). Small studies, careful writing, tools teams actually use.

Where I can help

Clinical decision support with calibrated risk and decision-curve thresholds.

Fast, safe demos on de-identified data with clear guardrails.

Wet-lab + analysis: reproducible figures and tidy code.

Writing support for abstracts, methods, model cards, and results narratives.

Project ideas

ED triage for suspected PAM (pilot)

Retrospective n≈30–50; quick feature audit.
Calibration curve, Brier score, decision-curve analysis.
Deliverables: brief report, model card, small demo UI.

Focused, ~4–6 weeks once data are ready.

Montenegro’s Medium — reproducibility note

Structured growth dataset with timing and counts.
Baselines with uncertainty notes; methods-ready figures.
Open data + open figures bundle (CSV + PNG).

Good fit for a letter / short communication.

How I work

Scope first
We draft the one-pager before we write code.
Evidence over hype
Every claim earns a figure. Every figure gets a caption.
Small, shippable units
Weekly milestones that stand on their own.

Recent outcomes

MM growth figures: 0–168 h curves, batch-freshness comparison, and passage-timing study with uncertainty bars.
Clinical ML demo: triage UI with reliability plot, conformal-style threshold option, and decision-curve view.

Have an idea that fits?

Send a short note and the smallest useful dataset.

Contact me →See a live demo