AI Engineer - Voice

Harper•San Francisco, CA

2d•Onsite

About The Position

We are building an AI-powered insurance brokerage that is transforming the commercial insurance market. Voice is our primary growth lever, as the majority of insurance transactions occur over phone calls. We are seeking an exceptional Head of Voice AI to architect voice systems that will power thousands of daily conversations across growth, sales, operations, and customer service. Initially, you will leverage low-code platforms like VAPI or Retell for rapid prototyping and deployment of production voice agents. As we scale, you will progressively build custom infrastructure using LiveKit and Pipecat. Your systems will manage a wide range of tasks, from cold outreach and lead qualification to complex multi-turn underwriting conversations and 24/7 customer support. Our commitment is to 'Staying REAL' with our AI systems, ensuring agents are Reliable, Experience-focused, Accurate, and have Low latency. You will collaborate directly with the CEO and CTO, with a strong emphasis on taking action. Our core principles include: 'There is no try, there is just do,' 'Actions lead to information, always default to action,' and 'Strong opinions lead to information.'

Requirements

5+ years of software engineering experience with at least 3 years deeply focused on voice/audio systems.
Production voice AI experience - you've built and deployed systems handling 10K+ minutes/month.
Hands-on experience with low-code voice platforms like VAPI or Retell for rapid prototyping.
Deep understanding of telephony and media protocols including SIP, RTP, and WebRTC.
Experience with voice orchestration frameworks like LiveKit, Pipecat, Daily, or custom-built solutions.
Advanced audio processing knowledge - you understand VAD, AEC, noise suppression at a technical level.
Proven ability to achieve sub-second latency in production voice systems.
Strong proficiency in Python and TypeScript/Node.js specifically for real-time systems.
Experience with both inbound and outbound calling at scale (1000+ concurrent calls).
Modern AI provider expertise with OpenAI, Anthropic, Deepgram, ElevenLabs, etc.
Track record of shipping voice products that directly impact business metrics.
Strong debugging skills for complex, multi-service voice pipelines.
Must be based in San Francisco and work in-office (relocation assistance provided).

Nice To Haves

You've built production voice AI systems handling 100K+ minutes per month with real customers.
You understand the full voice stack from telephony protocols to WebRTC-based media servers.
You can architect systems that maintain context through 30+ minute conversations.
You have experience with modern STT/TTS providers and know how to optimize them.
You ship voice features daily and iterate based on real conversation data.
You balance starting with practical solutions while building toward technical excellence.
You understand that voice AI is about business impact, not just technical sophistication.
You embrace "there is no try, there is just do" as your engineering mantra.

Responsibilities

Deploy production voice agents within weeks using VAPI or Retell for quick prototyping and immediate business impact.
Transition to custom voice infrastructure with LiveKit and Pipecat as volume scales.
Achieve sub-700ms latency across the entire voice pipeline while maintaining conversation quality.
Scale to 10,000+ concurrent calls with appropriate architecture evolution and optimization.
Integrate telephony at scale with Twilio, Telnyx, and enterprise SIP infrastructure.
Build outbound prospecting agents that identify qualified leads, overcome objections, and book appointments.
Create lead nurturing systems with personalized follow-ups that move prospects through the sales funnel.
Implement predictive dialing and call pacing algorithms for maximum efficiency.
Design qualification workflows that gather key information and route to appropriate human agents.
Develop form-filling agents handling 20-30 minute insurance application conversations.
Build underwriter follow-up systems that collect additional risk information through natural, multi-turn dialogue.
Create document collection workflows guiding customers through providing licenses, photos, and business documentation.
Implement intelligent escalation paths that know when to loop in human underwriters.
Design 24/7 policy servicing agents that explain coverage, generate certificates, and process endorsements.
Build claims intake systems that empathetically gather first notice of loss (FNOL) information.
Create payment processing agents handling failed payments, billing updates, and payment plans.
Develop proactive outreach systems for policy renewals, payment reminders, and important updates.
Create no-code/low-code tools enabling non-technical teams to create and modify voice workflows.
Build conversation analytics tracking quality metrics, completion rates, and customer satisfaction.
Develop A/B testing frameworks for voice personas, prompts, and conversation strategies.
Implement voice agent templates for common insurance workflows.
Create comprehensive monitoring to track latency, accuracy, and conversation outcomes.