Skip to content
Back to projects

Android & Backend Product

Voice-to-Notes

AI speech-to-text product serving 100K+ users — I build its backend workflows and Android surfaces.

Node.jsFirebaseCloud FirestoreAndroid

Type

Android & Backend Product

My role

Backend & Android contributor (Web3Task)

Ownership

company product

Result

100K+ users

Overview

Voice-to-Notes is a Web3Task AI product that turns speech into organized notes. I contribute to its backend and Android surfaces — it is company-owned.

Challenge

At 100K+ users, audio sessions, transcription state and notes must stay consistent across devices while long recordings process concurrently.

Results

Shipped features running in production for more than 100,000 users, with measurable reliability gains.

My Ownership

  • Orchestrated backend workflows for audio session management, transcription state tracking and concurrent processing
  • Synchronized notes, sessions and transcription state across devices with Firebase Realtime Database and Cloud Firestore — maintaining a <200ms sync-latency target in common flows
  • Investigated 30+ user-reported transcription issues through crash logs and backend traces, improving successful transcription flow reliability by ~8%
  • Built 5 Android recording widgets and quick-access flows (15–25% feature-adoption lift)

Technical Implementation

  • Async pipelines with explicit transcription state tracking so long audio survives retries and restarts
  • Firestore as the cross-device source of truth, with Realtime Database for low-latency sync paths

Engineering Challenges

At real user scale, edge cases — backgrounding, process death, slow networks — stop being edge cases.

Gallery