AI Engineer — Ad Creative Intelligence System (Claude + GPT)

I’m an in-house creative strategist running paid media for 3 brands (Meta, Google, TikTok — combined $1–5M/month spend). I know what makes ads win; I don’t know how to build the system that helps me find, produce, and scale winning creative faster. I want to build an “AI brain” — a working system, not a demo — that uses Claude and GPT together to help me do my actual job: analyze performance data, generate and refine ad concepts/copy, spot patterns in what’s winning vs. losing, and eventually help automate parts of testing and reporting. I’m looking for someone who will build the system for me and actively teach me how to use it to guarantee winning ads What you’ll • Design the LLM orchestration architecture (how Claude and GPT are used, when, and for what — not just “call the API”) • Build integrations with Meta Marketing API,, TikTok Ads API • Build the feedback loop that turns performance data into better future creative output • Set up a vector database / RAG system so the tool “remembers” brand voice and past winners • teach me how to use it for the 3 brands I work sith Required experience • Production experience building systems on top of Claude and/or GPT APIs (not just chatbot wrappers — actual orchestration, tool use, RAG) • Experience integrating with ad platform APIs (Meta Marketing API, or TikTok Ads API) — at least one required, more is a strong plus • Comfortable with Python, API integration, and basic data pipeline work • Can explain technical decisions in plain English to a non-engineer — this matters as much as the code • Portfolio or references showing something you built that’s actually in production, not just a prototype Nice to have • Experience in adtech, martech, or growth/performance marketing tooling specifically • Experience with vector databases (Pinecone, Weaviate, or similar) and RAG pipelines Red flags for me (please don’t apply if) • You’d propose “just fine-tune a model” as the first move — this needs sound architecture more than fine-tuning • You can’t point to anything shipped and running with real users/data Compensation Open to discussing compensation based on your experience and how we structure the engagement (hourly, milestone-based, or a mix).