Voice-agent specialist needed — tune + optimize an outbound lead-qualifying agent

**Title:** Vapi AI voice-agent specialist needed — tune + optimize an outbound lead-qualifying agent (ElevenLabs, low-latency) **Overview:** I need an experienced Vapi voice-AI specialist — someone who builds and tunes AI calling agents for a living — to optimize an outbound agent I've already set up for my home care agency. I have a separate developer handling the Make/CRM automation; YOUR focus is purely the AI agent: making it sound human, responsive, and reliable on real phone calls. **What's already in place:** - Vapi outbound assistant "Warm Lead Qualifier" (built and published) - ElevenLabs Flash v2.5 voice (Matilda) · Claude Haiku 4.5 model · Deepgram transcriber - Smart endpointing (LiveKit), low-latency settings, auto-fallbacks enabled - System prompt + first message written (qualifies warm inbound leads, books a callback, avoids health details on purpose) - Dedicated phone number provisioned **What I need from you (AI agent work only):** 1. Make it sound natural and human on REAL phone calls — right now it's slightly robotic/laggy in testing. Tune voice selection, stability/similarity, speed, and endpointing for natural turn-taking with no awkward pauses. 2. Optimize latency end-to-end (transcriber → model → voice) so responses feel real-time. 3. Refine the conversation flow / system prompt for smooth qualifying, graceful handling of interruptions, voicemail, and "bad time" responses. 4. Set up voicemail detection + an appropriate voicemail message. 5. Configure the end-of-call report so structured data (intent, timeline, callback details) is cleanly captured for my automation developer to consume. 6. Recommend best practices you've learned building these (filler words, background sound, barge-in handling, etc.). **Must have:** - Demonstrable experience building/tuning Vapi (or Retell/Bland) voice agents — please share examples or recordings of agents you've built - Deep familiarity with ElevenLabs voice settings and latency optimization - Understanding of LLM prompt design for voice (not just text chatbots) **Nice to have:** - Experience with regulated / sensitive industries (healthcare, home care, finance) where tone and compliance matter **To apply:** Share 1-2 examples of voice agents you've built (recordings ideal), describe how you approach latency/naturalness tuning, and give your estimated hours. Fixed-price preferred. This is the AI-agent half of a larger project — clean, well-documented work that my automation developer can plug into is what I'm after.