ANDROID, RESEARCH, DATA COLLECTION, SAMSUNG HEALTH

INSPIRE

Research Period: 2025.09 - Present

Overview

INSPIRE is a production-level research tracking app designed for real study operations. It is currently in internal testing and preparing for a full-scale study. The project focuses on reliable data collection under mobile OS constraints, repeatable configuration for different studies, and practical monitoring so researchers can verify data quality during deployment.

Context and Collaboration

INSPIRE is conducted in collaboration with Yonsei University College of Nursing. I owned the end-to-end engineering scope of the mobile app (excluding the cognitive games module).

Problem

  • In real studies, “it works on my phone” is not enough. Reliability must hold across devices, OS background limits, and daily routines.
  • Multi-source data collection increases operational risk: missed notifications, interrupted background work, recording failures, and inconsistent data coverage.
  • Researchers need visibility into whether data is actually being collected day by day, not just stored somewhere in the backend.

Goals

  • Collect multi-source behavioral and health data reliably in real-world conditions.
  • Support repeatable, study-specific configurations without rewriting the app.
  • Provide monitoring and researcher tooling to detect missing data early and reduce operational overhead.

Data Sources and Collection Scope

  • Sensor data: accelerometer and device context signals (e.g., phone state, screen state).
  • Health data (Samsung Health SDK): 6 key health data types aligned with the study protocol.
  • Surveys: scheduled 4 times per day.
  • Audio recordings: short, clip-based audio responses collected within the survey flow.

System Design (Operational Focus)

  • Config-driven study setup: most protocol settings are defined in JSON-based configurations, aligned with the modular tracking architecture direction.
  • Offline-first collection mindset: data is captured safely under unstable connectivity and synced reliably later.
  • Monitoring and observability: operational visibility is treated as a first-class requirement for real deployments.

What I Built / Contributed

  • Built the INSPIRE mobile app end-to-end (Android), covering data collection modules, scheduling, storage/sync, and user flows (excluding cognitive games).
  • Integrated the Samsung Health SDK and aligned health data with sensor and survey pipelines.
  • Implemented scheduled surveys (4/day) and audio recording collection flows.
  • Added monitoring using Firebase Crashlytics and structured logging to support debugging during internal testing.
  • Developed researcher tooling to:
    • download/export study data, and
    • check daily-level coverage (whether data arrived each day per participant), enabling faster issue detection during deployment.

Reliability Work (Implementation Details)

  • Hybrid scheduling for real-world constraints: I combined WorkManager + AlarmManager + ForegroundService to balance reliability and OS restrictions.
    • WorkManager runs periodic background sync workers (e.g., Health/Sensor/External metric sync) on a practical cadence suitable for OS constraints.
    • AlarmManager triggers time-critical reminders using exact alarms (with limited re-reminders).
    • ForegroundService supports continuous sensor collection and uses health-check style restarts to recover from interruptions.
  • Audio collection as an explicit user action: audio is recorded through a user-triggered survey interaction (start/stop button), producing short clip-based responses saved as .m4a (AAC).
  • Hybrid storage strategy (Firestore + Storage): structured study data (surveys, sensors, health metrics) syncs to Firestore, while audio files upload to Firebase Storage. The resulting download URL is stored back into the Firestore response metadata.

Tech Stack / Tools

Android, Kotlin, Jetpack Compose, Firebase (Firestore, Storage, Crashlytics), WorkManager, AlarmManager, ForegroundService, Samsung Health SDK, JSON-based configuration, researcher export/coverage tooling

Outcome

A production-level research tracking app ready for real deployment: multi-source data collection (sensors + Samsung Health + surveys + audio), operational monitoring via Crashlytics, and researcher-facing tooling for export and daily coverage verification. The system is designed to scale from internal testing to a full study with reduced operational risk.

Next

  • Expand the deployment workflow from internal testing to the full study.
  • Continue hardening reliability and document key lessons learned through blog posts.