Chefizi is an AI-powered culinary assistant available on iOS, Android, and web that helps chefs and home cooks create recipes, plan meals, and explore cuisines using natural language conversation. Powered by Claude AI, the app understands dietary restrictions, ingredient availability, cooking skill levels, and personal taste preferences to generate tailored recipes and cooking guidance in 21 languages. The platform includes a native mobile app built with React Native and Expo, a full-featured web application at ai.chefizi.app that mirrors all mobile capabilities, and a comprehensive admin panel for user management, content moderation, analytics, and a built-in blog CMS. Monetization is handled through a 5-tier subscription model with Apple In-App Purchases via RevenueCat, Stripe for web payments, and Robokassa for the Russian market — all with geo-based pricing that adjusts to local purchasing power.
!The Challenge
The culinary AI space is growing rapidly, but most existing solutions are simple recipe search engines wrapped in a chatbot interface. They lack true contextual understanding — they cannot remember a user's dietary restrictions across conversations, adapt to what ingredients are actually available, or generate genuinely creative recipes that respect cultural cooking traditions. Furthermore, none of them offered a truly multilingual experience that goes beyond basic translation. The client wanted to build a product that felt like having a personal chef consultant available at any time — one that remembers your preferences, understands when you say 'I have chicken and some vegetables,' and can walk you through cooking techniques step by step. The app needed to work across iOS, Android, and web with synchronized conversation history, support voice input for hands-free kitchen use, and be monetizable from day one with a subscription model that works globally. A critical technical challenge was achieving feature parity between the mobile app and the web application while maintaining a local-first architecture. Users expected their conversation history, saved recipes, and preferences to sync seamlessly between their phone and browser, even when offline. The subscription system needed to work across Apple IAP, Google Play Billing, Stripe, and Robokassa — each with different receipt validation flows and webhook patterns.
Our Approach
We architected the AI layer around Claude API with carefully engineered system prompts that establish the assistant's culinary expertise, personality, and boundaries. The AI maintains per-user memory that persists across conversations, storing dietary preferences, skill level, kitchen equipment, and taste profiles to provide increasingly personalized responses.
The mobile app was built with React Native and Expo SDK 54 for maximum cross-platform code sharing while maintaining native performance. We implemented local-first data storage with a 6-table sync schema that enables offline usage and bidirectional sync with the server via push/pull API endpoints.
Voice input was integrated using OpenAI's Whisper API, enabling hands-free recipe queries — essential for users who are actively cooking. The voice flow captures audio, transcribes it server-side, and feeds the text into the same AI pipeline as typed messages.
We designed a 5-tier subscription model (Trial, Home Cook, Professional, Business, Enterprise) with RevenueCat managing Apple IAP and Google Play subscriptions, Stripe handling web payments in most markets, and Robokassa serving the Russian market. Geo-based pricing adjusts subscription costs based on the user's country.
The web application at ai.chefizi.app was built as a standalone Next.js app sharing the same API backend, with a Claude.ai-inspired UI featuring a collapsible sidebar, conversation management, and full feature parity with the mobile app including voice input and file uploads.
The Solution
AI Engine & Memory System
The conversational AI is powered by Claude API with a custom system prompt that establishes deep culinary expertise across global cuisines. Each user has a persistent memory store that the AI reads at the start of every conversation, containing dietary restrictions, preferred cuisines, skill level, available equipment, and past recipe favorites. The AI can update this memory during conversations — for example, when a user mentions they recently went gluten-free. This creates a progressively more personalized experience that no generic recipe search can match. Streaming responses ensure minimal perceived latency.
Cross-Platform Mobile & Web Apps
The React Native mobile app runs on iOS and Android via Expo, with native-feeling navigation, haptic feedback, and platform-specific UI adaptations. The web app at ai.chefizi.app provides identical functionality in a browser with a professional desktop-optimized layout inspired by Claude.ai's interface — featuring a collapsible sidebar with conversation history, project organization, and settings. Both platforms share the same Express.js backend API and PostgreSQL database, with a local-first sync system using 6 database tables that enables offline access and conflict-free bidirectional synchronization.
Monetization & Global Payments
The subscription system supports five tiers with different feature limits (message counts, recipe saves, voice minutes). RevenueCat manages Apple IAP and Google Play subscriptions, handling receipt validation, subscription lifecycle events, and entitlement checks. Stripe processes web subscriptions for most countries, while Robokassa serves the Russian market. Geo-based pricing adjusts costs — for example, subscriptions in developing markets cost less than in the US or EU, with 21% IVA tax automatically applied for EU customers. All payment events sync to the backend via webhooks for unified subscription state management.
Admin Panel & Content Platform
The admin panel at admin.chefizi.app provides comprehensive platform management: user analytics dashboard, subscription metrics, individual user detail views with conversation history, referral program tracking, and a full legal documents editor with CKEditor integration. A built-in blog CMS powers 60+ articles across English and Russian, with SEO metadata management. The admin panel also includes marketing tools, consent management for GDPR compliance, and a cookie banner system supporting 15 languages. All legal pages (privacy policy, terms of service, acceptable use policy, etc.) are editable directly from the admin interface.
Results & Impact
Technology Stack
React Native / Expo
Cross-platform mobile app (iOS & Android)
Next.js
Web application at ai.chefizi.app
Express.js
Backend API server
PostgreSQL
Primary database with local-first sync
Claude API
AI-powered culinary conversation engine
Whisper API
Voice-to-text transcription for hands-free use
RevenueCat
Apple IAP & Google Play subscription management
Stripe
Web payment processing with geo-based pricing
Apple Sign In / Google OAuth
Social authentication providers