Keep the risk model inspectable
Deterministic aggregation and risk logic sit beneath any generated explanation so the primary signal can be tested independently.
MyHealth
An Android health companion that combines Health Connect data, quick habit logging and a local risk engine for recovery and burnout signals.
experiment · prototype
Product designer and builder · Updated Aug 2026
- Health data
- Android
- Risk modelling
- Personal analytics
01 · Problem
Health signals and daily habits live in different places, making it difficult to see when poor sleep, workload and recovery are converging into a meaningful risk pattern.
02 · System
The app synchronizes Health Connect data, combines it with quick habit logs, calculates daily risk summaries and schedules notifications for recovery and critical conditions.
03 · Outcome
The prototype includes biometric aggregation, tested risk logic, background synchronization, quick-log widgets and AI-assisted health insight generation.
Context
This is a personal experiment, not a medical device. Its value is reflective pattern recognition; it does not diagnose or replace professional care.
Architecture
- 01Health Connect data access
- 02Daily summary aggregator and risk engine
- 03Firebase-backed settings and persistence
- 04WorkManager synchronization and notification jobs
- 05Jetpack Compose dashboard, charts and quick-log controls
Decisions & trade-offs
Evidence
- Unit-tested risk engine
- Background health synchronization and daily summaries
- Quick logging for habits such as coffee and nicotine
What this taught me
- Generated interpretation is safer when it explains a visible, deterministic signal instead of creating the signal itself.