Vendor Diversity & Data Fragmentation
Each new wearable or telehealth platform required a completely bespoke integration — its own authentication flow, its own data model, its own storage schema. As the platform grew to support more clinical trial sponsors and diverse device requirements, this approach became unsustainable. Every vendor added weeks of development, duplicated code, and maintenance burden.
Vendor diversity problem: Apple Watch connects directly via HealthKit. TempTrack uses physical kits with multiple uniquely identified patches. Dexcom streams CGM data via webhooks. Every device had a completely different integration pattern.
Patient data fragmentation: With no common data layer, clinical trial investigators couldn't view patient health metrics across devices in a unified dashboard — defeating the purpose of passive monitoring.
Three-Layer Generic Architecture
Vendor Configuration
Super admins register any wearable or telehealth vendor (Apple, Fitbit, Oura, etc.) by entering API credentials. No code deployment required for new vendor onboarding.
Generic OAuth2 Engine
A single reusable authenticator handles the full OAuth2 flow — authorization request, token exchange, refresh — driven entirely by vendor config.
Dual-Table Data Architecture
A shared BigQuery profile table stores common columns, while each vendor maintains its own isolated table for unique or proprietary data fields.
Unified Pipeline
Incoming device data is normalized through a shared processing layer before storage. Raw payloads are transformed and mapped to a consistent schema.
System Architect — Designed & Built End-to-End
I personally designed this architecture from scratch. The challenge wasn't just technical — it required understanding clinical trial workflows (how investigators assign devices, how patients authorize, how study sponsors configure device requirements per protocol) and translating that into a flexible system that could absorb any new vendor without touching the core codebase.
I identified the repeating integration pattern across all vendor docs, abstracted it into a configurable model, and designed both the backend data pipeline and the BigQuery schema strategy. I led the implementation alongside the engineering team, defining the table structure, the OAuth2 engine design, and the data normalization contracts between layers.
Technologies Used
Measurable Results
- Reduced new vendor integration time from weeks of bespoke development down to a configuration + schema task — a structural shift in scalability.
- Clinical trial sponsors could now require any combination of wearable devices in a study protocol without triggering new engineering sprints.
- Patient health data from Apple Watch, Fitbit, Oura, Dexcom, and others converged into a single BigQuery-backed pipeline.
- The vendor-isolated table design in BigQuery preserved each platform's unique data schema while avoiding column pollution in the shared profile layer.
- Supported the shift to decentralized clinical trials (DCT) by enabling continuous passive patient monitoring from home.