The healthcare problems I care about,
and how I'm starting to solve them.
Three problems in healthcare, three working prototypes — neonatal monitoring, SaMD regulatory infrastructure, clinical evidence design. Built in code, grounded in published standards and clinical literature.
SpO2 pipeline · Babysat
SaMD-OS · 5 specialist agents
Pathway Designer · FHIR CarePlan
Nourish · flash_crash
380,000 preterm infants go home each year with no personalized monitoring.
Home monitoring fires constant alarms because it uses one-size-fits-all thresholds. Babysat builds personalized baselines from each infant's own data — and decides on physiology, not threshold crossings.
One-size-fits-all thresholds at home.
380,000 preterm infants discharge each year with BPD or CHD. Most go home with fixed SpO2 thresholds that don't account for gestational age. A 90% alarm on a 34-week preemie fires constantly.
Alarm fatigue hides the events that matter.
Preterm SpO2 ranges (92–95%) overlap with term "borderline" values (97–100%). Fixed thresholds either fire on every breath or miss real desaturation. The bottleneck isn't sensor accuracy — clinically significant desaturation depends on depth and duration, not threshold crossings, and home monitoring doesn't model either.
Personalize the baseline. Decide on physiology, not the threshold.
Babysat builds 14-day rolling baselines from each infant's own data, with GA-adjusted reference ranges validated against Castillo 2008 and Dawson 2010. Alarms fire only when both depth and duration cross significance — not on raw threshold dips.
Clinical narratives translate raw numbers into context parents can act on. Provider handoff reports include BISQ-R sleep assessment and tokenized viewer access for the next clinician on the chart.
From home monitoring to a clinical evidence stream.
Validate the personalized-baseline approach in a small NICU follow-up cohort against fixed-threshold practice. Wire HL7v2 interop into a real EHR pilot so triage flags surface inside the existing nursing workflow, not a separate app.
The arc: every infant's home monitoring data becomes part of their clinical record, so the next provider doesn't start blind — and population-level baselines refine themselves as more infants discharge.
SaMD teams can't ship because their regulatory artifacts are disconnected from their codebase.
SaMD ships continuously, but regulatory artifacts don't. SaMD-OS treats them as a build artifact — generated from code, reviewed by specialist agents, kept in sync with what regulators will actually see.
Docs run on a separate track from code.
SaMD ships continuously. Design controls in spreadsheets, risk files in Word, SOUP by hand. A 510(k) reflects six standards at once, drawn from sources never built to stay in sync.
Artifacts drift the day they're written.
By submission the codebase has moved. Teams that shipped first have zero traceability between what's in production and what regulators will see. Visible cost: delays and FDA AI requests. Hidden cost: QA leads burning out as the reconciliation layer.
Regulatory documentation as a build artifact, not a deliverable.
SaMD-OS generates IEC 62304 design controls, ISO 14971 hazard analyses, and SOUP registers from the codebase itself, then routes them through specialist reviewers — regulatory, QA, safety, cybersecurity, clinical — that mirror an FDA panel.
For teams already in production, three reverse-engineering paths recover what should already exist: code → SOUP, code → design inputs, code → hazard candidates.
Each reviewer is built from a canonical reference document — FDA guidance, an ISO standard, or published clinical literature — and pinned to a specific version so its judgment doesn't drift. Your team's own QA decisions and meeting notes feed back into the review context, so domain knowledge accumulates in the system, not in someone's head.
A regulatory CI/CD layer.
Pilot with SaMD teams approaching 510(k) to validate artifacts hold up to FDA submission end-to-end. Wire reviewers into the pull-request loop so every design-control update ships with the code that triggered it. Extend coverage from FDA to EU MDR / IVDR.
The longer arc: a regulatory CI/CD layer where every commit knows which design input it satisfies, which SOUP component it touches, and which hazard it might introduce — turning a six-month pre-submission scramble into something that's already done.
Zero connected strength training device studies in the literature — evidence exists but no product bridges it.
Resistance training is first-line therapy for sarcopenia. Supervised adherence is 72%, unsupervised 43%. The Clinical Pathway Designer closes the gap with a shared device data spine that keeps provider and patient on the same page between visits.
Resistance training works, but adherence collapses without supervision.
Resistance training is first-line therapy for sarcopenia (ICFSR 2018, strong recommendation). Supervised adherence runs 72%. Remove supervision — which insurance won't reimburse long-term — and it drops to 43%. Functional outcomes (gait speed, grip, TUG) respond strongly to progressive loading; muscle mass measures don't.
No connected device closes the loop between provider and patient.
Across 34 digital interventions for adults 60+, zero use a connected strength device. Apps coach, wearables count steps, but nothing surfaces actual session data — load, sets, compliance — to the provider. The bottleneck isn't efficacy; it's the silent eight months between supervised discharge and the next functional decline.
A shared device data spine for provider and patient.
The Clinical Pathway Designer connects provider workflow and patient journey through a shared device data spine — load progression, session compliance, baseline strength — visible to both sides at every visit. Goals are functional-only (gait speed, grip, TUG), no body composition theater.
Every phase of the pathway maps to a specific ICFSR adherence barrier — cost, transportation, lack of support. Seven evidence anchors (E1–E7) trace every design decision back to published literature. An illustrative FHIR CarePlan demonstrates EHR-shippable structure.
From design pathway to fielded clinical evidence.
Pilot with one geriatric clinic and one connected strength device manufacturer to validate the pathway end-to-end. Build the FHIR CarePlan integration into a real EHR so adherence data lives where the provider already looks.
The arc: a closed-loop sarcopenia care pathway where the device is part of the medical record, the provider sees adherence in chart review, and reimbursement follows function-restoration outcomes — not visit volume.