Caipher is a comprehensive mobile application I built to solve a real problem facing TikTok Shop affiliate creators: navigating the platform's strict violation policies and crafting appeals that actually win. As an affiliate creator myself, I experienced firsthand the frustration of receiving violation notices with no clear guidance on how to respond, watching legitimate sales evaporate while accounts sat in review purgatory. Caipher transforms that stressful experience into a structured, data-driven workflow powered by AI.

At the core of the app is an AI engine built on Anthropic's Claude that analyzes violations through either text descriptions or screenshot uploads. The system identifies the specific policy at play, scores the strength of a potential appeal on a one to ten scale, predicts realistic success rates as a percentage, and generates a professional appeal letter optimized to fit within TikTok Shop's strict 500 character submission limit. Each analysis follows a proven conversion framework with five strategic components: respectful opener, specific context, strongest defense, commitment to compliance, and polite request for reconsideration. The AI also recommends three to five specific supporting documents creators should attach to maximize their approval odds, tailored to the exact violation type.

Beyond appeal generation, Caipher tracks the complete lifecycle of every violation through a status workflow that moves from open to in appeal to won or lost. A submission helper copies the generated appeal to the clipboard and opens TikTok directly, eliminating friction between drafting and submitting. Creators can record outcome notes documenting what TikTok said and what they learned, building a personal knowledge base that informs future appeals. The dashboard surfaces key metrics including active violations requiring attention, appeal success rates, and overall account health at a glance.

Technically, Caipher is built as a cross platform React Native application using Expo SDK 54 with TypeScript throughout. The backend leverages Supabase for authentication, secure data storage with row level security policies, and screenshot uploads to private storage buckets. The AI layer integrates the Anthropic Claude API with custom prompt engineering, defensive JSON parsing, and multiple safety nets including character limit enforcement and dash stripping for output formatting consistency. The architecture uses a clean separation of concerns with dedicated service layers for violations, AI analysis, and storage, plus a state based navigation system and a custom component library that maintains a polished, brand consistent UI across the entire app.

Caipher represents the intersection of three things I care deeply about: building tools that solve actual problems for real users, leveraging AI thoughtfully to augment human judgment rather than replace it, and crafting mobile experiences that feel polished, fast, and genuinely useful. The app is currently in active development with plans to expand into brand collaboration management, unified messaging, smart reply templates, and additional creator workflow tools that complete the picture of a true all in one assistant for the modern affiliate creator.