AI Automation Engineer — Python + ComfyUI/Flux for Automated Fashion Content Pipeline

ABOUT THE PROJECT I run a faceless fashion brand on Instagram with a French, maximalist, editorial aesthetic. Think Vogue-level art direction. The account is growing and I'm ready to move from semi-manual workflows to a fully automated content system. I'm hiring one engineer to design and build that system end to end. This is a build project with potential for ongoing retainer work for the right person. CURRENT SYSTEM (WHAT YOU'D BE BUILDING ON) There's already a working Python automation running daily: Scheduler-based pipeline (schedule library) posting 2 photos + 1 reel per day Content currently sourced from a Pinterest board and a TikTok URL queue (Google Sheets) Captions auto-generated via the Claude API Posting to Instagram with video upload via Cloudinary Dedup tracking, retry/error handling, and email notifications on completion Your job is to migrate this pipeline from sourced content to original AI-generated content: the scheduler, captioning, and posting layers stay; the content source becomes a ComfyUI/Flux generation pipeline (with Pinterest demoted to a style-reference input via img2img rather than a repost source). You'll inherit clean, working code. WHAT YOU'LL BUILD A fully automated Instagram content pipeline: Image generation. ComfyUI/Flux pipeline producing original editorial fashion imagery in a consistent house aesthetic, including LoRA-trained recurring AI models (consistent faces across all content) Reference-driven generation. image-to-image / style-transfer workflows that take reference imagery (Pinterest boards, moodboards) as aesthetic input and output original generations in that direction Automated captioning. Captions and hashtags via LLM API (Claude/GPT) in a defined brand voice Scheduled posting. Automated publishing via the official Instagram Graph API only (no third-party bots or methods that risk account flags) Performance feedback loop. Automated pull of engagement data via API, scoring of post performance, and a mechanism that feeds learnings back into generation parameters so the system self-optimizes Review queue. A lightweight approval step so I can QA generations before posting, with the option to phase out later Required Skills Python. strong, demonstrated experience; this is the backbone of the entire system ComfyUI / Stable Diffusion / Flux — portfolio must show character consistency work (same face/model across many images via LoRA or equivalent) API integration. Instagram Graph API, plus AI APIs (Replicate, fal.ai, OpenAI, Anthropic) Automation infrastructure. Schedulers, queues, error handling; the system must run reliably without daily babysitting Data analysis. Pull and analyze post performance data (pandas or similar) and translate it into pipeline adjustments Nice to Have AI video generation (Kling, Runway) for Reels Experience with fashion/editorial/aesthetic-driven content specifically Simple dashboard experience (Streamlit or similar) for the review queue and performance reporting How to Apply In your proposal, answer these two questions specifically, generic proposals will be skipped: 1- What is your current preferred image generation stack and why? (Be specific about models and tools.) 2- Link to portfolio work showing the same AI character/model across multiple images. Hiring Process Shortlisted candidates will be offered a small paid test task: generate a 6-image editorial set with one consistent AI model matching a reference aesthetic I'll provide. The build contract goes to whoever delivers the strongest test result.