Hands-on Building · AI & Agentic Workflows

AI Fitness OS: one energy-balance dashboard to replace a dozen health apps

A personal fitness operating system that consolidates fragmented health data — calories, activity, steps, workouts, weight — into a single, unified view. Built end-to-end as a non-technical founder, with AI doing the technical lift.

Role
Product owner & builder
Type
Full-stack web app
Approach
AI-assisted, spec-driven
Stage
Live · in configuration
The Problem

Health data lives everywhere — and nowhere useful

Steps in one app, calories in another, workouts in a third, weight in a fourth. The single number that actually matters — energy in versus energy out — never existed in one place. AI Fitness OS was built to close that gap.

Unified dashboard

A single energy-balance view that pulls every source together, replacing multiple disconnected apps.

Apple Health sync

Steps, activity and workouts flow in automatically through a webhook ingestion endpoint — no manual entry.

Photo calorie analysis

Snap a meal and an AI vision model estimates its calories — logging food without hunting through a database.

How It Was Done

A spec-first build, executed with AI as the engineer

I brought the product vision and research; AI handled architecture, code generation and documentation. The work moved from idea to a running application through a tight, repeatable loop.

1

Define the outcome

Nailed the core job-to-be-done — one energy-balance number — before writing a line of code, so every feature laddered back to it.

2

Scaffold the full stack

Generated a complete Next.js + Supabase application, delivered as a downloadable package with a step-by-step setup guide.

3

Wire the data pipelines

Built the Apple Health webhook and the OpenAI Vision flow for food photos, with row-level security on all stored data.

4

Configure & iterate

Moved into live setup — resolving environment, keys and integrations — treating each hiccup as configuration, not failure.

The Build

Modern, production-grade foundations

Chosen for speed, security and a clean path to deployment — the kind of stack a lean product ships on.

Next.js 14 · frontend & backend
Supabase · Postgres with row-level security
Vercel · deployment
OpenAI Vision · food photo calorie analysis
Apple Health · webhook ingestion
Vitest · testing
Git · version control
Claude Code · development environment
The Outcome

From scattered apps to a working, unified fitness OS

A fully scaffolded, deployable application now exists and runs — the fragmented-data problem solved by design. It proves a non-technical operator can ship a real full-stack product by pairing clear product thinking with AI execution.

1
Unified energy-balance view
5+
Data sources consolidated
100%
AI-assisted, spec-driven build
Current state: The app is live and in the configuration phase. The photo-calorie feature is being finalised — an OpenAI Vision setup step — and a technical write-up of the build has already been drafted for public sharing.