Native AI Customer Support App

I need a fully native mobile application that can answer incoming phone calls on our main business number, speak to the caller in either a male or female voice, and resolve their questions from a built-in knowledge base—no external APIs allowed at any stage.

Core workflow
• A user logs in with their mobile number and a one-time OTP.
• Up to four additional numbers can be added; the login number remains the primary line.
• During first-time setup I’ll enter our business overview and a list of product features. This data becomes the searchable knowledge base the AI will reference when speaking with callers.
• When someone rings the primary line the AI greets them as our assistant, listens, searches the stored details and replies in the chosen voice. All exchanges are saved as a chat transcript that only the primary user can see.
• If the answer is not found, the AI politely offers to connect the caller to a senior. It then dials the registered numbers one after another until someone picks up, stopping the chain as soon as a call is answered.

Key requirements
– Completely offline / on-device AI inference—no cloud or third-party APIs for speech, language or telephony.
– Dual-voice TTS (male & female) with smooth switching.
– Multilingual interaction: the first release must handle English and Hindi; the architecture should let us plug in more languages later.
– Clean, native codebases (Kotlin/Java for Android and, if you cover iOS, Swift/Obj-C).
– Local data storage for the knowledge base and chat history (SQLite or similar).
– Robust telephony integration to pick up, speak, listen, and transfer calls seamlessly.

Acceptance criteria
1. Install, register, and add numbers without crashes.
2. AI answers calls, responds from the saved knowledge, and logs every exchange.
3. Escalation sequence dials registered numbers in order and stops on first connection.
4. Voice switching and bilingual replies work as configured.
5. All features function with airplane mode on (except the actual call, of course), confirming no external services are hit for AI or voice.

Hand-off deliverables
• Source code with build instructions.
• Brief technical document outlining the on-device models, speech engines, and call handling approach.
• A short video demo proving the five acceptance points above.

Looking forward to collaborating with someone who knows their way around on-device ML, TTS/STT engines, and native telephony APIs.