I need an experienced developer to embed an AI-powered voice assistant into my existing CRM web app. Scope • Capture microphone input directly in modern browsers (no desktop plugins). • Convert speech to text reliably, then pass the transcript to a large-language model for intent detection. • Map each recognized intent to concrete CRM actions such as filtering records, creating new entries, or navigating to specific dashboard sections. • Return the result to the user via natural-sounding text-to-speech. • All processing must run in real time, keeping latency low enough for a smooth conversation. Key Requirements • Works inside a CRM (category: CRM (Customer Relationship Management)). • Secure authentication so only logged-in users invoke commands. • Clean, well-documented JavaScript/TypeScript code that plugs into the existing frontend framework (React). • Server-side components (e.g., Node.js, Python) set up to call chosen STT, LLM, and TTS services—OpenAI Whisper / GPT / ElevenLabs or similarly accurate alternatives. • Intent-to-action routing table that is easy for me to extend. • Graceful fallbacks when audio permission is denied or the model cannot interpret a request. Deliverables 1. Frontend module for microphone capture, streaming STT, and playback. 2. Backend endpoints and middleware for intent processing and command execution. 3. Setup instructions and environment variables template. 4. Short demo video showing the assistant creating an entry, filtering data, and jumping between CRM sections by voice. 5. One round of post-delivery tweaks based on real-world testing feedback. If you have built browser-based voice interfaces or LLM-driven assistants before, I’m eager to see examples.