DNP-prepared nurse. I build clinical measurement tools.
CV and resume available on request.
I am a DNP-prepared nurse and a self-taught software engineer. Three and a half years of pediatric critical care and burn nursing at Children's of Alabama showed me how much clinical work runs on tools that do not exist yet, so I started building them. Now most of my attention is on medical imaging: measuring pediatric brain tumors the same way every time, so the decision to treat a child rests on something steadier than a hand-drawn line.
Before building a measurement tool, I wanted to know what the field already does. How is the size of an optic pathway glioma actually measured in the published literature, how much of that work is volumetric rather than two-dimensional, and how often does anyone check whether two readers measuring the same tumor agree?
A scoping review of 28 studies published between 2008 and 2025, sole author, following PRISMA-ScR. I characterized how each study measured tumor size and whether it reported reliability. Five of the 28, about 18 percent, reported inter-rater or intra-rater reliability at all.
PRISMA-ScR protocol across PubMed, Embase, Cochrane, Web of Science, and IEEE Xplore. Studies were sorted into manual, semi-automated, fully automated, and approximation-formula measurement categories, then charted for reliability reporting, imaging sequences used, and how growth was defined.
In revision for submission. The finding that most of this literature does not report measurement agreement is the gap the segmentation study was designed around, which is why these two sit next to each other.
When a child is badly burned, two numbers shape the first day of care: how much of the body is burned, and how much fluid that child needs. Both get worked out by hand, quickly, often in the middle of the night.
Burn resuscitation runs on two calculations made under pressure: what percentage of the body is burned, and how much fluid that child needs in the first 24 hours. Get the surface area wrong and the fluid volume is wrong with it. Nurses learn this from a paper chart.
Total body surface area feeds the Parkland formula directly, so surface area error propagates linearly into fluid volume. Pediatric body proportions differ from adult ones and shift with age, which is why the rule of nines fails in children and Lund-Browder exists. The chart is static paper and the arithmetic is done under time pressure.
A burn assessment trainer for phones. You map the burn on a body diagram, it does the math, and it explains what it did. English or Spanish, and it works with no signal.
A trainer that walks through Lund-Browder surface area mapping and Parkland fluid calculation, then checks the result against burn center referral criteria. It runs offline on a phone, in English or Spanish, with an eight-chapter teaching curriculum in both languages.
Offline-first progressive web app covering age-adjusted Lund-Browder mapping, Parkland calculation, burn center referral criteria, and an eight-chapter curriculum, fully localized to Latin American Spanish (es-419) across both interface and teaching content. Shipped as a PWA on the web and wrapped for iOS.
Age-adjusted body maps handle the surface area math, and the app shows its work at every step instead of just handing back a number.
An offline-first progressive web app, wrapped for iOS. Age-adjusted Lund-Browder tables drive the surface area math, and a teaching engine explains each step rather than returning a bare number. More than 1,700 automated tests cover the calculation paths.
React 19 and TypeScript on Vite, Tailwind for styling, Zustand for state, Workbox for the service worker, Tauri 2 for the iOS and desktop wrappers. Locally stored session data is encrypted at rest with AES-256-GCM. More than 1,700 automated tests under Vitest and Playwright, weighted toward the calculation paths, because a wrong fluid rate is the kind of error that matters.
On the App Store since March 2026, and free on the web.
Live on the App Store since March 2026 and on the web; the current version is 1.5.1. It is labeled an educational tool and does not claim to support clinical decisions at the bedside.
Released March 18, 2026; version 1.5.1 August 4, 2026; published under Martin Apps LLC. Educational use only. The app makes no clinical decision support claim and carries that disclaimer in-product, which is a deliberate constraint on what it is allowed to say rather than a legal footnote.
Board prep question banks tell you an answer was wrong. They rarely tell you why your reasoning was wrong, and the reasoning is the part that decides the next question you miss.
A nurse practitioner board prep app pairing an adaptive question bank with a conversational tutor you can argue with. It tracks which topics you keep missing, schedules them back on a spaced-repetition interval, and estimates pass probability from your own performance rather than a generic curve.
React Native on Expo with Supabase behind it, and a large language model driving the tutor. The adaptive engine weights re-drills toward repeated misses, which is the study method I use myself: master the questions, then teach them back.
Live on the App Store as BoardBound Prep since June 2026; the current version is 1.0.4.
Wilderness reference information is worth the least at exactly the moment you need it most, because that is usually somewhere with no signal.
An offline-first survival reference drawn from US Army FM 21-76 and federal sources including CDC, EPA, NOAA, and the National Weather Service, with environmental calculators and plant identification. The whole corpus ships inside the app. Nothing in it requires a connection.
React and TypeScript as a progressive web app, with the reference content stored in IndexedDB behind a Workbox service worker, wrapped for iOS. The content is sourced and attributed rather than written by me, because survival guidance is not a place to improvise.
Live on the App Store as Wilderness Survival Companion since March 2026; the current version is 1.8.0.
I had automated coding capacity going unused overnight, and no safe way to point it at real work without risking the parts of my projects that should never change unattended. Clinical calculation code is the obvious example.
A task runner that turns unused capacity into bounded, reversible, pre-approved work across several projects. Its distinguishing feature is a safety gate: any automated change touching clinical domain logic gets an independent review from a different model family before it can land, and a critical finding reverts the commit rather than warning about it.
Node.js running a five-phase pipeline on a schedule. Projects are flagged for clinical gating individually, and a flagged project routes any change under its domain or data directories to a cross-family review step. The gate fails closed: a critical finding reverts the commit and logs the outcome.
In daily use across my own projects. The source is private.