Overview
At Lupa, we're hiring a Forward Deployed Engineer for an AI-native luxury hospitality operator — a lean, high-output team running 60+ premium short-term rental units across 13 Class-A buildings, with an expansion roadmap to 7+ U.S. cities.
This is not a strategy seat and it is not a demo seat. MIT studied 300 corporate AI projects and found 95% delivered no measurable return — not because the models were weak, but because nobody watched the process before automating it. Your job is to be the person who watches first.
You'll go where the work actually happens — a housekeeping crew running 300+ turnovers a month, a customer service pod answering guests seven days a week, people coding invoices by hand, a revenue team moving rates across a portfolio — map how it really works, then decide step by step what should be plain software, what should be AI, and what should stay with a person because being wrong is expensive. Then you build it into the tools the team already uses, and you stay on it until people are voluntarily using it after you walk away.
If you're energized by total scope, a direct line to the CEO, and a portfolio of real deployments you'd never get to own at a bigger company, this is a rare seat.
What You'll Do
Run the discovery. Sit with a housekeeper, a CS rep, and the person who codes invoices. Map the real process, not the documented one. The undocumented rule nobody wrote down is the whole game — and you only find it by watching.
Make the split. Assign every step on purpose: fixed rules stay ordinary software, messy judgment goes to the model, expensive mistakes stay with a person. Knowing what not to automate is core to the job.
Build into the real stack. Ship real code into the tools already in use — Guesty, PriceLabs, Breezeway, HubSpot, ClickUp, QuickBooks, Supabase, Google Workspace, and existing n8n automations. No greenfield.
Design for failure. Retries, human gates, fail-closed on money and guest data. Build evals from real historical cases with known-correct answers so you can count and read the misses.
Drive adoption. The part most people skip and the reason most of this work fails. A tool that ships but isn't used removes zero hours. Budget more time for trust than for building.
Own the number. Verified human hours removed per month, from workflows still in use 30 days after handoff. You set the baseline before you build — including building measurement where none exists yet (e.g., 14 housekeepers running 300+ turnovers with no timeclock).
Who You Are
Must-haves
You've deployed a workflow inside someone else's business and stayed with it until people actually used it. You can bring the before-number and the after-number.
You can code: Python or TypeScript, plus SQL. You debug your own stack trace. This is not an advisory role.
You've integrated into a stack you didn't choose — APIs, webhooks, auth, other people's data models.
You've run LLM systems in production: prompting, structured output, tool use, agent loops, retrieval, and a real understanding of how models fail.
You've broken something in production and fixed it. We'll ask how you found out and how bad it got first.
You've had an adoption failure — something you built that people rejected or quietly abandoned — and you can tell us what you changed. If you've never had one, you've never been close enough to the work.
You use AI in your daily workflow, out loud and disclosed (Claude Code or equivalent).
Fluent written and spoken English.
You'll go where the work is: on-site in Chicago and on calls with the Bogotá team, watching people work. You treat a housekeeper or support rep with respect — they know things about this business you never will, and they only tell people who respect them.
Nice-to-haves
Working Spanish. Two of the three highest-volume teams operate in Spanish.
Experience in a two-sided marketplace or B2B platform environment.
This is not for you if: you're a consultant whose last deliverable was a slide deck; you need a greenfield; you automate everything on reflex; or you'd ever say "the business didn't adopt it" instead of "I missed something."
What's In It For You
What we can actually offer: total scope, a direct line to the CEO, and a portfolio of live deployments across multiple departments you would not get to own anywhere with a bigger budget.
Growth: you'll own end-to-end systems from discovery to production to adoption, in a company scaling from one city to 7+ — with your work directly driving units-growing-faster-than-headcount.
Structure: distributed team across Chicago and Latin America. Fast decisions, high standard, meritocracy, radical real-time feedback.
Process note: this role starts with a paid deployment trial before a full-time offer, in both directions — you get to see how we operate, and we've learned the hard way that a great interview isn't proof.
Apply through Lupa and we'll guide you through every step.