Push-to-talk platform, next generation
A field-operations voice platform was running its audio over a TCP message broker. I audited the existing eight-component estate — Go, Android, Angular, more than a hundred endpoints — and designed the replacement.
- Evaluated the real options head to head: LiveKit against Mumble, NATS against RabbitMQ, SeaweedFS against MinIO.
- Specified the migration to a WebRTC SFU on LiveKit, with QUIC/WebTransport primary and WebSocket fallback.
- Wrote the specifications for floor control, multi-channel PTT, priority interruption, secure sessions, and voice recording.
- Consolidated several backend services into one coherent platform architecture.