<div class="content-intro"><p data-pm-slice="1 1 []">Groupon connects 42 million customers with local experiences - restaurants, events, wellness and travel - and the million-plus merchants who deliver them. Our mission is to get people offline and into the real world at prices that make it possible. We're an AI-native company in the middle of a platform transformation, moving from a deals marketplace to an experience discovery platform that works for customers and merchants at the same time.</p> <p> </p></div><p>Every deal Groupon sells passes through one internal platform — Groupon IQ — where deals are created, checked and published. Seven engineers run it, and AI agents write much of the code. It's absorbing the company's other admin tools one by one; a deal lookup that used to take two minutes now takes thirty seconds. Nobody yet owns its reliability, or decides when agent-written code is safe to ship. You'd own both — and still be in the code.</p> <h2>What you'd actually get to build</h2> <p>Groupon IQ already runs on the company's modern engineering foundation — the platform the rest of the org is still migrating onto. That's your advantage: you build on the finished thing while making it load-bearing for everyone.</p> <ul> <li><strong>One TypeScript codebase</strong> — ~95% of it, in a single repository, on the framework built by ex-Netflix engineers (Encore).</li> <li><strong>Infra without tickets</strong> — declare a database, cache, or queue in a line of code and it's provisioned, wired, and observed for you: 60+ services stood up without a single infra request.</li> <li><strong>Ship all day</strong> — every branch is a full staging environment; one pull request can change 23 services and two frontends and go out as a single deploy. Merges reach production the same day — minutes when it matters.</li> <li><strong>A workflow engine (Temporal)</strong> runs the long processes — purchase flow, deal creation, outreach — reliably and auditably.</li> <li><strong>AI is native</strong> — a shared model gateway and coding-assistant rules wired into every service. Engineers here average ~7 hours of real coding a day (versus ~3 elsewhere at Groupon); agents write real code, and review keeps it honest.</li> </ul> <h2>Who you'll work with</h2> <p>You'll report to Pavel Jakl (B2B Tribe Director), who writes code himself — a <strong>player-coach</strong> seat, not a calendar-and-1:1s one. You'll lead ~7 engineers across Prague and Bangalore in one codebase. Here, <strong>reliability is the product</strong>: as IQ absorbs more of the company's tools, more teams depend on it not breaking. Not for you if you want to stop coding, or want a pre-scoped backlog with someone else on the pager. Right for you if you like making a load-bearing platform faster and more reliable as it scales — and deciding how humans and agents build on it.</p> <h2>A week in the life</h2> <ul> <li><strong>Backend reliability, end to end</strong> — a long-running workflow stalls silently; you trace it through the platform's observability (no ticket, no waiting), fix it in one pull request, and write the post-mortem that stops it recurring.</li> <li><strong>Production code, daily</strong> — real output, not a title.</li> <li><strong>The AI code-review bar</strong> — recently an agent traced intermittent database timeouts to a call that never waited for its job to finish, moved every path onto the shared polling client, and added the tests that were missing: agent-written, human-reviewed, shipped. You own the standard that keeps work like that safe to merge.</li> <li><strong>Absorbing the next tool</strong> — IQ is consolidating the company's admin tools; you make each one land without dragging the whole platform down.</li> <li><strong>Cross-team architecture calls</strong> — you make them and defend them.</li> <li><strong>Mentoring</strong> — seven engineers toward end-to-end ownership; feedback within a day, standards written down.</li> </ul> <h2>Signals you're a fit</h2> <ul> <li>There's been a week where your commits and your team's output were both on the line — and both shipped.</li> <li>You've owned a platform where reliability <em>was</em> the product, and you moved the incident number.</li> <li>You've reviewed agent-written code and decided where the merge bar sits.</li> <li>You've traced a silent workflow failure end to end and written the post-mortem that stopped it recurring.</li> <li>You've automated away a manual task and spent the freed time building the next thing.</li> </ul> <p>We'll dig into these live — bring the specifics. <em>(You're still shipping, so show us something you built recently where agents wrote much of the code, and how you kept it correct.)</em></p> <h2>What success looks like</h2> <ul> <li><strong>Day 0–30:</strong> met everyone, mapped the incident history, shipped a real backend fix, started instrumenting a reliability baseline (there isn't a trustworthy one today).</li> <li><strong>Day 31–60:</strong> the baseline is trustworthy; the AI code-review standard is live; one named slice of reliability debt closed.</li> <li><strong>Day 61–90:</strong> incident rate measurably down against your baseline; an agent runs continuously, taking a manual task off the team; the next admin-tool migration lands without a reliability hit.</li> </ul> <h2>How we operate<