I’m building a 24/7 voice-based, first-level technical support line for my IoT product and I want the experience to feel completely human. When a customer in India rings our toll-free 1800 number, the agent must: • Answer naturally in either English or Hindi, switching seamlessly when the caller does. • Identify the customer’s issue through free-form conversation, guided probing, or a predefined question flow. • Consult a pre-built knowledge-base (I’ll supply the content) and return the most relevant fix or troubleshooting steps in real time. • Decide—based on confidence and predefined rules—whether to: – resolve the case on the call, – create and email/SMS a ticket automatically, or – warm-transfer the caller to a live engineer with full context. Technical expectations • Python-based solution so my team can extend it. • Integrates with a standard SIP trunk or Twilio-like voice API to sit behind our 1800 number. • Uses reliable ASR, NLU, and TTS to keep latency low and speech lifelike. • Modular workflow engine where I can tweak intents, responses, and escalation logic without touching core code. • Logs every interaction (audio + transcript) and pushes summaries to our existing ticket system (REST API available). Deliverables 1. Working voice agent running in a staging environment. 2. Clean, commented Python code with virtual-env or Docker setup. 3. Documentation covering deployment, knowledge-base updates, and escalation rule edits. 4. One round of post-deployment tuning after a week of live traffic. If you’ve built conversational IVRs, voice bots, or Twilio Flex/Voiceflow/Dialogflow CX projects, I’d love to see examples. Let’s bring a human touch to automated tech support together.