Health OS

Your body.
Better observed.

A personal experiment in translating wearable data, clinical research, and nutrition context into a more legible model of one body over time.

View Live Demo

Demo password

vital-atlas-52-meridian

The story

It started as a question about interpretation.

I switched from an Apple Watch to a Garmin. The hardware and training signals were useful, but the product made it hard to reason across them.

So I started building a different interface. That small design problem became a research problem: how much of the clinical and exercise science literature can be made practical when it is connected to daily wearable data, food intake, and recovery patterns?

Health OS is my working answer. It does not claim to diagnose anything. It tries to organise the evidence I already have into a clearer model: what changed, what might explain it, and what I should pay attention to next.

40+
Clinical metrics tracked
6
Organ systems modelled
Open
Research sources reviewed
1
Subject in the first study

The useful question is not whether health software can be personalised. It is whether we any longer need to wait for it.

Most health apps are not designed for people whose bodies, conditions, or goals fall within the average. They work well when your needs fit a predefined category, but they start to break down when your health story is more specific, more chronic, more complex, or simply not commercially common enough to be deeply served.

Health OS is a small counterexample: one person, one data history, one set of goals, and a research layer that tries to make the interpretation explicit for acheiving specific & personalized goals.

The broader idea is that people can now build personal health ecosystems around problems that may never look commercially large enough for a big technology company. Patients and physicians dealing with chronic conditions, including long neglected conditions like PCOS, can create tools that track what matters to their own biology and daily life.

That does not replace clinical judgment. It changes the starting point. Instead of waiting for a generic app to decide a condition is worth serving, people can assemble data, research, and AI into systems that expose reasoning, show uncertainty, and adapt to the biological question in front of them.

AI is useful here only if it improves interpretation. The clinical evidence still matters. The wearable data still needs structure. The product problem is connecting those layers without turning ambiguity into false confidence.

Feature 01

A live map of
your entire body.

Garmin provides fragments: strain, sleep, HRV, training load, and muscle activity. Health OS groups those signals into a body level view so patterns are easier to inspect.

HeartStable

HRV and resting pulse are read as a recovery context, not isolated scores.

LungsImproving

VO2 max and oxygen trends sit beside training history.

LiverReview

Nutrition timing is flagged as a hypothesis, not a conclusion.

LegsLoaded

Recent training volume is separated from readiness language.

CoreSparse

Low direct stimulus is treated as missing evidence.

Upper BodyRecovered

Recovery estimates are shown with the data that produced them.

Feature 02

Nutrition as
structured evidence.

Instead of treating food as a diary entry, the system parses plain language meals into approximate nutritional evidence that can be compared with training, recovery, and micronutrient needs.

You said
"Had a bowl of oats with banana and peanut butter this morning, grilled chicken salad for lunch, and a whey protein shake after the gym."
Health OS extracted
Protein142g
Carbohydrates198g
Healthy Fats28g
Dietary Fibre18g

Feature 03

The metrics your
apps rarely connect.

The interesting work is not another dashboard. It is deciding which research backed metrics are useful enough to operationalise, how uncertain each signal is, and how they should influence daily decisions.

Healthspan Score

A composite view of biological systems, useful mainly as a way to track direction rather than declare truth.

VO2 Max Trajectory

Current VO2 max is placed inside a longer trend with training load, recovery, and nutrition context.

Muscle Recovery Model

Detected muscle work is translated into a simple recovery ledger so training decisions have memory.

Stress Load Index

HRV, sleep quality, and recovery load are combined into a signal that is meant to prompt investigation.

Micronutrient Tracking

Repeated nutrition gaps become visible over time instead of disappearing inside daily food logs.

AI Analysis Layer

The AI layer is used to summarise competing signals and surface hypotheses, not to pretend certainty.

Try it yourself

The live demo is open.

Built on my own data and shared as a working prototype for people interested in health interfaces, applied AI, and quantified self systems.

Open Health OSView GitHub

Demo password

vital-atlas-52-meridian