Senior Backend Engineer, GenAI

What you're applying for

Anyone can ship an AI agent. Shipping one that’s right ~98% of the time, with no human checking its work, takes different engineering. That’s what this team builds.

We're looking for a Senior Backend Engineer to help push our AI customer-service agents closer to fully autonomous insurance service. You'd be joining the team behind the agent that talks to real customers every day over email, phone, chat, and SMS - handling policy questions, claims conversations, and verified documents end-to-end, with no human in the loop. This isn't a demo or a copilot. Real customers, real money, real stakes, every day.

In this role you'll

• Design, build, and own production LLM agent systems end-to-end - deciding how the agent thinks, what it's allowed to do, and how it fails safely
• Work on our real-time voice agent, where the hard problems are conversational: interruptions, mid-sentence tool calls, and keeping response times imperceptibly fast
• Shape the next generation of our knowledge base - how the agent's knowledge is structured, curated, and served, so answers stay grounded as coverage grows
• Ship changes as controlled experiments, measure them against thousands of replayed real conversations on our in-house eval platform, and graduate winners into production
• Track cost per conversation across millions of LLM calls and find the wins that cut spend without cutting quality
• Own what the model knows and says - the customer context it sees, the limits on what it may reveal, and the guardrails against confident nonsense
• Build and operate scalable backend services in TypeScript/Node on Kubernetes/AWS with deep observability - if you ship it, you own it

What you'll need

• 6+ years building production-grade backend systems that real users depend on
• Hands-on experience shipping LLM-powered features to production, including honest reckoning with non-determinism, hallucination, and cost
• An evaluation mindset - you don't call a change "better" without a measurement, and you're comfortable with the statistics that back that claim
• A track record of debugging distributed systems in production, and the discipline to find root cause across service boundaries
• Fluency with agentic coding tools like Claude Code or Cursor, including the habit of treating AI-generated code with a senior reviewer's skepticism
• The ability to pull your own numbers, design your own experiments, and measure your own impact
• Ready to work in an office environment most days of the week