AI Automation Engineer (E-commerce)

About the role
We are looking for a freelance AI Automation Engineer to build and run an agent system that operates our Shopify dropshipping business 24/7, with minimal human input.
We are a European e-commerce company with 7 years in Shopify dropshipping and more than 250 stores built. We run fashion stores for Europe (France, Belgium, the Netherlands, Poland, Denmark and more) and for the US and Canada. Today a human team handles the daily chain: product research, listing, creatives and ad launches. We want AI agents to take over most of that chain.
You live in the world of agentic AI. You know Claude, ChatGPT, OpenClaw and Hermes Agent inside out, and you have shipped agents that run on a server without babysitting. You design and build the system and hand it over fully working. One human controller on our side reviews and approves the key decisions.

• Type: freelance gig, project-based
• Location: fully remote
• Hours: flexible, with some overlap in the evening (CET)

The mission
Build five agents that run our daily chain on a server, around the clock.
Each agent hands its output to the next. The one fixed human step is a single approval before money is spent on ads.
What you will build
You own the system end to end, across five stages.
1. Product research
◦ Connect agents to Winninghunter (MCP/API) and scan daily for winning products, scaling ads and competitor stores.
◦ Score products on clear criteria: ad longevity, engagement, number of stores selling it, margin and shipping.
◦ Deliver a daily shortlist with the reasoning behind each pick.
2. Product launch
◦ Create products and collections in Shopify automatically via the Admin API or Shopify MCP.
◦ Generate product photos with AI (e.g. Higgsfield), plus titles, descriptions, variants and pricing.
◦ Write copy in the language of each market (French, Dutch, Polish, Danish, English).
◦ Apply the right conversion strategy: offers, bundles, upsells, social proof and page structure.
3. Creatives
◦ Turn competitor winning ads into new concepts: hooks, angles and offers.
◦ Design static and video ad creatives with AI image and video tools, in all needed formats.
◦ Write headlines and ad copy per market.
4. Ad launch
◦ Launch campaigns through the Meta Marketing API (Meta Ads MCP), with fixed naming, budgets and testing structure.
◦ Build in approval gates and hard spend limits, so no money goes out without the rules allowing it.
5. Deep analysis
◦ Pull ad metrics and Shopify orders daily and calculate real profit per product and per ad.
◦ Apply kill and scale rules, spot creative fatigue and suggest next tests.
◦ Feed the learnings back into research and creatives, so the system gets better every week.
Across all stages you also handle orchestration, logging, monitoring, error handling, API cost control, key security and clear documentation.
Tech stack
This is what we use today; you are free to propose better choices.
Area
Tools
AI models
Claude (API, Claude Code, Agent SDK), OpenAI / ChatGPT
Agent frameworks
OpenClaw, Hermes Agent, MCP servers
Product research
Winninghunter
Store
Shopify (Admin GraphQL API, Shopify MCP)
Creatives
Higgsfield and other AI image/video tools
Ads
Meta Ads (Marketing API / MCP), later TikTok Ads
Infrastructure
Linux server (VPS or cloud), Docker, Python and/or Node.js
Shared workspace
Google Drive and Google Sheets
Must-haves

• At least 2 years of hands-on work with LLM APIs, including agent systems running in production. You can show them.
• Deep knowledge of the Claude and OpenAI APIs: tool use, function calling, MCP, prompt design and context management.
• Hands-on experience with OpenClaw and/or Hermes Agent, and you can explain when to use which.
• Strong Python and/or TypeScript, plus REST and GraphQL APIs and webhooks.
• Linux server skills: Docker, process managers, cron or queues, logging, monitoring and backups for 24/7 uptime.
• Experience with the Shopify Admin API and the Meta Marketing API.
• You understand e-commerce, not just code: what makes a winning product, what converts, and Meta metrics like CTR, CPM, CPC, CPA and ROAS.
• A security mindset: secrets management, scoped permissions and spend limits.
• Clear written English. You document everything you build.
• You work independently and proactively, without daily hand-holding.

Nice to have

• You have run a dropshipping store or bought Meta ads yourself.
• Experience with AI image and video tools such as Higgsfield, Midjourney, Runway or Kling.
• Experience with spy tools like Winninghunter, Minea or PiPiADS.
• Copywriting skills in French, Dutch, Polish or Danish.
• Experience with the TikTok Ads API.
• Experience building evals to test prompt and agent quality.
• You can build simple dashboards for reporting and approvals.

Project milestones
Within 2 weeks, the full chain from research to analysis runs on the server, with a human only approving the key steps.
Phase
Deliverable
Done when
Milestone 1 (days 1–4)
Server live; research agent and product launch agent working
A daily product shortlist arrives automatically, and approved products go live in Shopify with photos and copy
Milestone 2 (days 5–9)
Creative agent and ad launch agent working
Creatives and campaigns are built automatically and launch after one approval, within spend limits
Milestone 3 (days 10–14)
Analysis agent and feedback loop working; documentation done
Daily profit reports, kill/scale suggestions and learnings flow back into research and creatives
What we offer

• A clear project with real ownership: you design the architecture.
• A competitive project fee, based on experience.
• Full access to the tools and API budget you need.
• Direct contact with the owner: short lines, fast decisions.
• A real business with real data, so you see the impact of your work right away.
• Fully remote, flexible hours.

How to apply
Send your application with the word AUTOPILOT in the subject line. Applications without it are not read.
Include:
1. Your CV or LinkedIn profile.
2. A short video (max 5 minutes, e.g. Loom) walking through an agent system you built and that runs in production.
3. Answers to these questions (max 1 page):
◦ How would you architect the five-stage pipeline above? Which tools would you use where?
◦ Which steps would you keep human-in-the-loop, and why?
◦ What would you estimate the monthly running cost to be (server plus API usage)?
◦ How do you prevent an agent from wasting ad budget when something goes wrong?
4. Your project quote (fixed price or hourly rate), estimated timeline and availability.