Full-Stack Developer to Build an AI Headshot Generator Website

Hi, I need a full-stack developer to build an AI Headshot SaaS. We are using WordPress.com Business with a PremiumPress theme foundation for our frontend shell, user profiles, and design layout, completely bypassing WooCommerce. All heavy AI loops, face splits, and conveyor belt mechanics must be offloaded to an isolated external backend (e.g., Node.js/Vercel or Supabase) to prevent WordPress server timeouts and crashes. I have a complete 35-page blueprint detailing all layouts, math distribution loops, and ready-to-use AI prompts so there is zero guesswork. OUR CONTRACT RULES: 1. Custom Hybrid Bridge: The frontend UI, O*NET search, and user dashboard will live in a custom WordPress template/plugin hooking into PremiumPress. The backend AI queue and image processing loops must run on an external serverless microservice. 2. NO WooCommerce: Using WooCommerce is permanently banned. Checkout must use a custom Paddle.js integration directly into custom database tables. 3. Live Testing Site: Build must happen on a private staging link. I must be able to view daily progress and personally verify functionality before milestone payments are released. 4. Daily Backups: All code must be pushed daily to my private GitHub account. 5. No Shortcuts: If you hit a technical theme limitation, you must explain it to me before making changes. 6. 90-Day Warranty: Include a 3-month warranty after Milestone 5 to fix backend bugs or API glitches for free. What is your locked hourly rate for future updates? DEVELOPMENT MILESTONES: * M1: Custom Tables & Search: Isolated database tables for uploads, queue jobs, and gallery states. Secure local import of 10,000+ O*NET professions data for instant autocomplete using server-side AJAX fetching. * M2: Frontend UI (Pages 1–5): High-fidelity, mobile-responsive conversion of wireframe Pages 1-5, layout styles, packages, and multiple selfie upload pipeline. * M3: Validation & Checkout: Integrate Amazon Rekognition API for selfie checks (blurry, closed eyes) AND automated celebrity upload block. Custom Paddle.js with payment webhooks. * M4: Backend AI Core: Asynchronous worker queue connecting to DeepInfra/Replicate (FLUX-2-pro API). Code loops mapping 50/50 face-lock IP-Adapter conditioning strength (1.0 vs 0.85 scale rules) and sequential conveyor belt rotation across 8 themes and 3 attires, saving final assets to a secure Cloudflare R2 bucket (zero egress fees). MUST include a "Mock API Environment toggle" in the backend to allow full infrastructure stress-testing (simulating 100+ concurrent user checkouts and generation loops using open-source tools like Locust or Artillery) without executing live, paid external API generation calls. * M5: Gallery, Email & Admin: Page 9 gallery with progressive loading, 4K upscale trigger, passwordless Magic Links, automated 30-day cloud data deletion script, SendGrid/Mailgun transaction email integration, Freshdesk footer widget snippet, and admin dashboard. CORE TECHNICAL QUESTION: Since this platform connects to APIs asynchronously, what happens if the API fails on image 60 of a 100-image order after a customer is charged? How do you write your code to handle partial failure and retries smoothly? *** PROJECT READY FOR DEVELOPMENT *** I have a comprehensive, 35-page detailed technical blueprint mapping out all frontend layouts, database keys, percentage-based "conveyor belt" loops, and optimized FLUX 2 PRO prompt arrays. There is zero creative guesswork required. Short preview pages of this master documentation will be shared with shortlisted engineering candidates during the initial text interview process to verify architectural compatibility.