Featured project
QWR AI Voice Bot
- Production AI voice agent with ~350–500ms time to first audio, built by decoupling the telephony layer (Exotel) from the AI engine (Gemini Live API).
- Bidirectional audio streaming with native barge-in (interruption handling) over WebSockets.
- Caller identity extraction via async LLM function calling, with post-call summaries delivered by a Telegram bot.
- Django Channels
- Gemini Live API
- Exotel WebSockets
- PostgreSQL / Supabase
- Telegram API