AI Automation Engineer
Build LLM-powered workflows for marketing operations — in the client’s own accounts, with the evaluation sets that decide whether they are good enough to ship.
Paid task, never free work · Feedback either way · 2–3 weeks end to end
About this job
This is applied engineering, not research. The workflows you build run unattended on real client operations — lead qualification, content pipelines, reporting, WhatsApp triage — which means they have to handle messy input, fail safely, log everything and stay cheap to run. Almost everything is built inside the client’s own accounts and API keys rather than ours, so the work has to be legible to somebody else afterwards. This role covers our US clients across time zones, which means overlap hours rather than an offshore handoff. It is the most competitive attention market we work in, accessibility exposure is a genuine legal risk rather than a nicety, and the rules change state by state.
What you’ll do
- Design and build agents and workflows against real client processes
- Write evaluation sets from real examples and score every agent before it ships
- Design the guardrails — approval steps, confidence thresholds, off switches
- Integrate with CRMs, forms, inboxes, WhatsApp Business API, analytics and CMSs
- Keep per-call cost down through sensible model selection rather than defaulting to the largest
What we’re looking for
- 3+ years building software, with real production LLM work among it
- Strong Python or JavaScript, comfortable with APIs, queues and webhooks
- You have shipped something that ran unattended and dealt with it breaking
- Able to explain your architecture to a non-technical client in their language
- Able to work United States client hours from wherever you are based
Nice to have, not required
- Experience self-hosting open-weight models for privacy-sensitive workloads
- Familiarity with n8n, Make or similar as well as hand-written services
- Any marketing operations background at all — it is rarer than it should be
You’d probably be unhappy here if…
- You want to do research or model training rather than applied delivery
- You would rather not talk to clients about what you built
- You believe most things should be fully autonomous on day one
We put this on every job for the same reason we put it on every service page: finding out in month three costs us both far more than reading it now.
What you get
- A published salary range — we do not negotiate people down from it, and we will not ask what you currently earn.
- Remote — United States — we ask for overlap hours with the team and your clients, not a chair.
- Paid learning time — courses, conferences and certifications covered, with time in the week to use them.
- Your own tooling budget, including AI tooling. Nobody here does a job badly because a licence was refused.
- A portfolio you keep — with client permission, the work is yours to show.
- A capped client load — we limit how many accounts we take on precisely so nobody carries twelve.
How to apply
Send us something you have built — a site, an audit, a campaign, a workflow, a piece of writing — along with a short note about why this role. A CV on its own tells us very little, and we would rather read one real thing than three pages of bullet points.
Put “AI Automation Engineer” in the subject line. We read every application and we reply either way, usually with a reason. If we go quiet for more than a week, chase us — that is our failure, not yours.
No recruiters or agencies, please. We hire directly and always have.
Other open roles
Not quite right? See everything we’re hiring for — or write to us anyway, which is how several people here started.
Back to all careers