After implementing a 30-Day Activity Calendar in EdgeFit, I wanted to move beyond simply displaying fitness metrics.
The next milestone was Personal Activity Insights.
EdgeFit is my privacy-first fitness companion built using React Native, TypeScript, HealthKit, and SQLite.
What's new?
Activity Consistency
Understand how many days have recorded activity across a recent period.
Weekly Step Comparison
Compare average daily steps across consecutive weeks, while accounting for missing data.
Exercise Frequency
Understand how many different days contain recorded workouts.
The engineering challenge
One important lesson when working with health data is that missing data is not the same as zero activity.
If a metric is unavailable, treating it as zero could produce misleading comparisons and insights.
That's why I'm focusing on explainable calculations and data-coverage requirements.
Why not use AI immediately?
I want to explore on-device machine learning, but first I need a deterministic baseline.
Rule-based insights are easier to test, explain, and evaluate.
Later, I can compare them with a small on-device model to determine whether machine learning offers measurable improvements.
Privacy-first architecture
EdgeFit builds on HealthKit and locally persisted activity data. Personal Activity Insights extends this existing architecture without requiring a cloud AI service.
What's next?
My next goal is to experiment with on-device ML and compare its usefulness, latency, memory consumption, and energy usage against deterministic insights.
GitHub: https://github.com/raviseta/edgefit
I'd love feedback from other mobile developers working with HealthKit, privacy-first architecture, or on-device intelligence.
Tags: reactnative, ios, privacy, healthtech
