Software Engineer

We build AI agents for hospitality.Our agents handle guest messages, phone calls, check-ins, marketing, and increasingly the operational work that used to require dedicated teams at every hotel. They run on top of legacy systems most of the industry has been stuck with for decades, and they actually do the work — not summarize it, not suggest it, do it.What you'll work onWe're hiring senior engineers who want to build that.You'll work on the agent stack itself — the orchestration layer that decides what an agent does, the skills that give agents specific capabilities, the integrations that let agents act on real systems, and the platform that ties it all together.Concretely, the work looks like:Agent orchestration. When a guest texts asking for early check-in, the agent has to look up the reservation, check housekeeping status, find an upsell, process payment, and update the system of record. You'll build the systems that route, plan, and execute that work reliably.New skills. Each new agent capability is a meaningful product surface — voice, marketing, operations, sales. You'll own one of these end-to-end with a small team.Integrations. Hospitality runs on systems that are decades old and certifications that take months. You'll build the connectors that make legacy systems usable by modern agents.Evals and reliability. Agents that work 95% of the time aren't useful at our scale. You'll build the testing infrastructure that lets us ship confidently.Who you areYou've been a software engineer for at least 6 years. You've shipped production systems that real users depend on, and you understand what it takes to keep them running.You love the idea of building AI agents. Not chatbots, not copilots, not "AI features." Agents that do the work. You've probably been building things on the side with Claude or GPT, and you have opinions about what's possible right now versus what's six months away.You write code with AI tools and you've gotten good at it. Claude Code, Cursor, whatever your stack is — you treat these tools the way you'd treat a strong collaborator: useful for some things, wrong about others, faster overall when used well. We don't care which tools you use; we care that you've internalized how to ship faster with them.