I’m looking for a confident freelancer to build an AI automation system based on OpenClaw for Amazon Featured Image workflows. You do not need to be highly experienced or senior. What matters is that you understand the problem and are confident you can build a working system. This is not a simple chatbot. The system must act as an agent-style AI that automates everything except the human graphic designer. Core Idea OpenClaw is the central AI agent. Google Drive is not just storage — it is the AI’s knowledge base and working context. The AI must: Access only permitted Drive folders Understand folder contents dynamically Handle empty folders at the beginning React when new files appear later (event-driven) There will be no dashboards. The system is controlled by chatting with OpenClaw. Telegram Architecture (Very Important) Telegram is the only interface, with two chat contexts: 1) Private Control Chat (Owner ↔ OpenClaw) Used for commands such as: “Start the relabelling process for folder X” “Generate draft images using cheap models” “Summarize everything done so far” “Check if any unresolved issues remain” This chat is private and used for orchestration. 2) Production Group (Owner + Designer + OpenClaw) Designer uploads PNG files OpenClaw automatically reviews them OpenClaw sends structured feedback Owner can intervene at any time Telegram is the UI. Workflow Phase 1 – Knowledge Processing OpenClaw reads Drive files (documents, screenshots, notes) and extracts: Approved copy Rules and constraints Risks Phase 2 – Image Draft Generation (Automated) Draft stage (cheap & iterative): gpt-image-1-mini Nano Banana Pro (Gemini image generation) Used for layout, color, typography direction, and copy placement. Optional polish stage (limited): gpt-image-1.5 Only for top 3–5 drafts Images are saved back to Drive. Phase 3 – Human Design Designer creates PSD manually Exports PNG Uploads PNG to the Production Telegram group Designer PNGs do not exist at the beginning. The system must wait and react when they appear. Phase 4 – Image Review & Feedback When a PNG is uploaded, OpenClaw must: Generate 100 / 150 / 200 px thumbnails Run OCR-based readability tests Measure: font size (px) line-height spacing alignment Detect typos vs approved copy Flag Amazon policy risks Compare with previous feedback Phase 5 – Feedback via Telegram (English) Feedback must be clear and actionable, e.g.: “Increase headline font size by +20%” “Move logo left by -8px” “Line-height too tight, add +3px” Structured as: MUST FIX IMPROVE MEASUREMENTS NEXT CHECK Each revision must confirm what is fixed or still missing. Technical Expectations Comfort with: Python Google Drive API Telegram Bot API OpenAI APIs (image + vision) OCR (Tesseract / PaddleOCR / EasyOCR)