On-Call IT/Support Engineer for Deployed AI Apps

About us Arkitekt is a custom software and AI automation agency. We build and host production AI systems for our clients — RAG pipelines, voice agents, automation workflows, and agentic systems — on an ongoing managed basis. That means live deployments that need maintenance, the occasional fix, and someone sharp who can jump in when a ticket comes through. The role This is an ad-hoc, hourly engagement. Work is ticket-driven and the load varies — some weeks are quiet, others are heavy when something breaks or a client needs a change. You won't be building greenfield projects; you'll be keeping deployed ones healthy. Day to day, you'd: Log into our deployment node and triage/resolve incoming IT and support tickets Implement fixes, small features, and config changes on already-deployed client projects Diagnose issues across the stack (app logic, integrations, APIs, hosting) and ship the fix Move fast using AI coding tools, while knowing exactly what the code you're shipping does What we're looking for We lean on AI coding tools heavily (Cursor, Claude Code, and similar) and expect you to as well — but as a force multiplier, not a crutch. You should be able to read what these tools generate, catch when it's wrong, and ship something you'd stake your name on. We're hiring an engineer who uses vibe coding to work quickly, not someone who just prompts and pastes. You're a fit if you have: Real software engineering / IT background — you understand systems, not just prompts Working knowledge of how modern AI systems are built and deployed (LLM apps, APIs, integrations) Comfort operating in someone else's codebase and infrastructure Strong debugging instincts and the judgment to know when AI output is plausible-but-wrong Reliable, responsive, and able to flex with a variable workload Nice to have Experience with RAG, voice/agent frameworks, automation platforms, cloud hosting/deployment, and on-call or managed-services work. To apply Tell us (briefly) about a time you used an AI coding tool to fix something in a live system — and one time you caught the tool getting it wrong before it shipped. Skip the generic cover letter; we read for signal.