MU
MUHAMMAD USMANSOFTWARE ENGINEER • FLUTTER & AI
BACK TO ALL PROJECTS
Phebsoft (Pvt.) Ltd.Lead Mobile & Audio Architect2024 – Present

Melogen AI Music

Generative AI-Powered Music Production & Streaming Platform

Audio Latency<120ms
PlatformsiOS & Android
Release Cycles8+ Prod Builds
Rating4.9 ★
THE ENGINEERING CHALLENGE

Integrating high-throughput audio generation models with seamless background audio playback and waveform visualization on both iOS and Android created severe memory overhead and audio thread stuttering.

THE ARCHITECTURAL SOLUTION

Engineered a dedicated Flutter audio pipeline utilizing custom platform channels, FFmpeg audio transcoding, and background service daemons. Designed a Python Flask API to handle asynchronous AI generation jobs with Redis queuing and S3 CDN delivery.

SYSTEM ARCHITECTURE & DATA FLOW

Asynchronous Audio Synthesis & Streaming Pipeline

High-performance media pipeline routing user prompts through GPU cluster endpoints and delivering real-time streaming buffers to mobile clients.

01 // STAGEFlutter Mobile Client (Audio Engine & Waveform)
02 // STAGEFlask API Gateway & JWT Auth
03 // STAGEGPU AI Music Generation Service
04 // STAGEFFmpeg Stream Chunking & S3 Storage
05 // STAGELow-latency HLS Playback & Cache Engine
KEY ENGINEERING HIGHLIGHTS & SOLVED BOTTLENECKS

Custom FFmpeg Playback & Caching Engine

Developed background audio caching and byte-range streaming to eliminate stutter on low-bandwidth networks.

Flutter • FFmpeg • Dart isolates

In-App Subscriptions & Token Credit System

Implemented RevenueCat and native App Store / Google Play billing with automated webhook reconciliation in Flask.

StoreKit • Google Play Billing • Flask

Waveform Scrubbing & Real-time Visualizer

Custom canvas painter rendering 60fps audio frequencies with sub-frame seek accuracy.

CustomPainter • Web Audio/FFT
TECHNOLOGY STACK & INFRASTRUCTURE
FlutterDartPython FlaskFirebase AuthFFmpegAWS S3DockerRevenueCat