This thesis presents the design, implementation, and evaluation of Yourang, an AIpowered voice assistant platform for automated business telephone communication. The platform enables companies across multiple industries—including restaurants, hotels, and service providers—to deploy intelligent voice agents capable of handling customer calls, processing reservations, and managing business operations without human intervention. The system integrates OpenAI’s Realtime API for natural language processing with Twilio and Telnyx VoIP infrastructure for telephony, establishing a real-time bidirectional audio pipeline. The organization-based architecture supports isolated environments for each business client, while a modular tool registry pattern enables extensible agent capabilities. Key features include bidirectional calendar synchronization with Google Calendar, Apple iCloud, and Calendly; a comprehensive internationalization framework with structured prompt engineering; and automated SMS notifications via BulkGate. The platform evolved from a proof-of-concept to a production-ready system through three development phases, employing clean architecture principles, dependency injection, eventdriven processing, and provider abstraction patterns. Performance evaluation demonstrates sub-second API response times and end-to-end call latency of approximately 1.2 seconds. The resulting system is positioned for commercial deployment, supporting multilingual voice interactions with automated customer communication and real-time calendar management.
Questa tesi presenta la progettazione, l’implementazione e la valutazione di Yourang, una piattaforma di assistente vocale basata sull’intelligenza artificiale per la comunicazione telefonica aziendale automatizzata. La piattaforma consente alle aziende di diversi settori—tra cui ristoranti, hotel e fornitori di servizi—di implementare agenti vocali intelligenti in grado di gestire le chiamate dei clienti, elaborare prenotazioni e gestire le operazioni aziendali senza intervento umano. Il sistema integra l’API Realtime di OpenAI per l’elaborazione del linguaggio naturale con l’infrastruttura VoIP di Twilio e Telnyx per la telefonia, stabilendo una pipeline audio bidirezionale in tempo reale. L’architettura basata sulle organizzazioni supporta ambienti isolati per ogni cliente aziendale, mentre un pattern modulare di registro degli strumenti consente capacità estensibili degli agenti. Le funzionalità principali includono la sincronizzazione bidirezionale del calendario con Google Calendar, Apple iCloud e Calendly; un framework completo di internazionalizzazione con prompt engineering strutturato; e notifiche SMS automatiche tramite BulkGate. La piattaforma si è evoluta da un proof-of-concept a un sistema pronto per la produzione attraverso tre fasi di sviluppo, impiegando principi di architettura pulita, iniezione delle dipendenze, elaborazione guidata dagli eventi e pattern di astrazione dei provider. La valutazione delle prestazioni dimostra tempi di risposta API inferiori al secondo e una latenza end-to-end delle chiamate di circa 1,2 secondi.
Design and implementation of yourang, an AI-powered voice assistant platform for automated business communication
KASHFIPOUR, SEYEDKIAN
2025/2026
Abstract
This thesis presents the design, implementation, and evaluation of Yourang, an AIpowered voice assistant platform for automated business telephone communication. The platform enables companies across multiple industries—including restaurants, hotels, and service providers—to deploy intelligent voice agents capable of handling customer calls, processing reservations, and managing business operations without human intervention. The system integrates OpenAI’s Realtime API for natural language processing with Twilio and Telnyx VoIP infrastructure for telephony, establishing a real-time bidirectional audio pipeline. The organization-based architecture supports isolated environments for each business client, while a modular tool registry pattern enables extensible agent capabilities. Key features include bidirectional calendar synchronization with Google Calendar, Apple iCloud, and Calendly; a comprehensive internationalization framework with structured prompt engineering; and automated SMS notifications via BulkGate. The platform evolved from a proof-of-concept to a production-ready system through three development phases, employing clean architecture principles, dependency injection, eventdriven processing, and provider abstraction patterns. Performance evaluation demonstrates sub-second API response times and end-to-end call latency of approximately 1.2 seconds. The resulting system is positioned for commercial deployment, supporting multilingual voice interactions with automated customer communication and real-time calendar management.| File | Dimensione | Formato | |
|---|---|---|---|
|
Kian_Kashfipour_10976831.pdf
accessibile in internet per tutti
Descrizione: Thesis Kian Kashfipour
Dimensione
2.98 MB
Formato
Adobe PDF
|
2.98 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/252487