Senior AI Voice Engineer — Conversational AI / Voice Agents

We’re looking for an exceptional AI Voice Engineer who has already built production AI phone agents that sound remarkably human.

The goal is simple:

Build an AI phone agent where the conversation feels so natural, fast, and fluid that it feels dramatically better than typical AI calling platforms.

We are specifically interested in a stack involving:

Grok / xAI + ElevenLabs + Twilio/SIP + custom integrations

This is not a basic chatbot project.

We care obsessively about:

• Extremely realistic human voice
• Very low conversational latency
• Natural interruptions / barge-in
• Knowing when the caller has actually finished speaking
• Natural pauses and pacing
• Short, conversational responses instead of robotic paragraphs
• Emotion, tone, and conversational variation
• Handling people who interrupt, change their minds, hesitate, or speak unclearly
• Function calling and real-world actions during calls
• Call transfers
• Scheduling appointments
• CRM integration
• Full transcripts, recordings, outcomes, and analytics
• Reliable outbound and inbound calling at scale

What You’ll Build

A production-quality AI calling system where:

Caller speaks → speech is understood → Grok determines the response/action → ElevenLabs generates extremely realistic speech → response happens naturally over the phone.

The agent should be able to:

• Have genuine back-and-forth conversations
• Interrupt and be interrupted naturally
• Remember context throughout the call
• Ask intelligent follow-up questions
• Handle objections
• Schedule meetings
• Transfer qualified callers to a human
• Access information through APIs
• Update our CRM automatically
• Trigger workflows after calls
• Classify call outcomes
• Generate summaries and next actions

We want the system engineered for real production use, not just a demo.

You Should Already Have Experience With

We strongly prefer someone who has personally built and deployed AI voice agents using several of the following:

• ElevenLabs Conversational AI / ElevenLabs Agents
• ElevenLabs real-time voice models
• xAI / Grok APIs
• OpenAI-compatible LLM APIs
• Twilio Voice
• SIP / telephony infrastructure
• WebSockets / streaming audio
• Speech-to-text
• Text-to-speech
• VAD / turn detection
• Interruption / barge-in handling
• Function calling / tool use
• CRM integrations
• Calendly / Google Calendar integrations
• Webhooks
• Call recording and transcription
• Real-time latency optimization
• Prompt engineering specifically for spoken conversations

Experience with systems such as Vapi, Retell, Bland, LiveKit, Pipecat, Deepgram, Cartesia, or similar platforms is useful — but we are looking for someone capable of going beyond an off-the-shelf implementation.

VERY IMPORTANT

Please do not apply if your experience is primarily:

• Chatbots
• Basic OpenAI API integrations
• Simple ElevenLabs text-to-speech
• No-code AI automations
• Voice-agent demos that were never used in production

We specifically want someone who can say:

“Yes, I have already built real AI phone agents like this.”

When Applying

Send us:

1. A recording or phone number of the best AI voice agent you have personally built.
2. Explain the exact stack you used.
3. Tell us approximately what the end-to-end conversational latency was.
4. Explain how you handled:

• interruptions
• turn detection
• silence
• background noise
• voi cemail
• call transfers

5. Tell us what you would use today to build the most human-sounding AI phone agent possible and why.
6. Tell us specifically what you have built using:
ElevenLabs, Twilio, Grok/xAI, Vapi, Retell, LiveKit, or similar technology.
7. Put the words “HUMAN VOICE” at the beginning of your application so we know you read the full posting.

What Success Looks Like

We should be able to call the finished system and think:

“This is completely different from the normal AI phone bots I’ve heard.”

Fast.

Warm.

Natural.

Intelligent.

Interruptible.

Not scripted.

Not robotic.

Not full of awkward pauses.

We will judge the work primarily by how the actual phone call feels, not by how impressive the architecture diagram looks.

If you have already built something at this level, send us the call recording. We want to hear it.