AI Agent Engineer — Paid Build Task First

AI Agent Engineer (Agentic Ecommerce Agency) We are building an agentic marketplace management company — TikTok Shop first, Amazon and Shopify behind it. Affiliate program management, listing CRO, creator outreach, GMV Max ads, violation handling. Most of that work is deterministic, and we intend to have agents run it end to end from a single onboarding call, with a small number of senior humans (TTS, media buying) supervising rather than executing. We are hiring one person to own that agent stack: design it, build it, run it in production, and train the team on it. How this hire works We are not hiring off a résumé. The first contract is a paid build task — $1,200, one week, real scope from our actual business. If it goes well, it converts to a long-term role at $65–$125/hr, and we have no ceiling for someone who is clearly worth more. If it doesn't, you keep the money and we part ways cleanly. The build task Given a TikTok Shop product listing URL plus a short brand brief, produce a scored CRO audit and a rewritten listing package — title, bullets, image brief, keyword targets — with the reasoning trail attached. It has to run unattended from that single input. We will score it on: Reliability — does it produce a usable output every time, or only when the inputs are clean? Failure behavior — what happens when the page won't load, the model returns garbage, or a step times out at 3am. Silent failure is a fail. Observability — can we see what it did and why, without reading logs by hand? Handoff — can a non-engineer on our team run it, read the output, and trust it? Cleverness is not on that list. What you'd own after that The agent architecture for client onboarding through full-service delivery — proposal, onboarding intake, listing CRO, creator outreach and follow-up, reporting Production operation: orchestration, state, retries, evals, monitoring, cost control Integrations, including MCP servers for anything not natively on the build path Defining where an agent decides and where a human approves — this is client-facing revenue work and the line matters Training our team to operate what you build, so it doesn't live only in your head What we need to see in your application A link or Loom of an agent system you have built that runs unattended in production. What it replaced, and what happened to the numbers after it shipped. Your actual stack, named: orchestration, models, memory and state, eval approach, monitoring. Your read on how you'd approach the build task above. Two paragraphs is plenty. We are looking for how you think about failure, not a proposal document. Applications without a real system to point at will not be reviewed. Requirements Shipped and operated agent systems in production, not demos or one-off automations Strong Python, comfortable with APIs, scraping, queues, and the parts of this that are plain engineering Model-agnostic. We use Grok/xAI and expect you to have opinions about where it wins and where it doesn't — but if you only know one provider, this isn't the role. Experience with ecommerce or marketplaces (TikTok Shop, Amazon, Shopify) is a strong plus, not a hard requirement Available for real overlap with US Central time Not a fit if Your experience is no-code workflow builders only You would describe an agent as "done" before it has run for a month without you watching it You need fully specified tickets — this is a build-from-intent role Long-term, full-time-equivalent, starting immediately. End of description. AI Agent Development · Python · API Integration · Automation · LLM Prompt Engineering · Model Context Protocol (MCP) · Workflow Automation · Ecommerce · Shopify · Marketing Automation