Platewise
Custom AI-powered recipe app that helps British home cooks plan meals in seconds, not hours
One search, four AI-powered recipe styles to choose from. See how OpenAI, Anthropic, Google, and XAI each interpret your ingredients differently. (Click to enlarge)
The Challenge
You're staring at the fridge wondering "what's for dinner?" for the third time this week. Most recipe apps use American measurements and ingredients, making them frustrating for UK cooks. Meal planning feels like a chore that takes hours every week.
Beyond the measurement conversion hassle, US-centric apps assume ingredient availability that doesn't match UK supermarkets, use unfamiliar terminology (zucchini vs courgette), and provide oven temperatures in Fahrenheit. For British home cooks, this creates friction at every step of the cooking process.
The Solution I Built
A web application where you enter your available ingredients, dietary needs, and preferences. Within 30 seconds, you receive four different personalized recipe suggestions—all using British measurements (grams, ml, °C), UK ingredient names (courgette, not zucchini), and references to UK supermarkets.
Beyond just recipes, the app includes a complete meal planning experience:
- Weekly meal planner with calendar view for organizing breakfast, lunch, and dinner
- Automatic shopping list generation that aggregates ingredients across your planned meals
- Pantry staples tracker so you can exclude items you always have on hand
- Smart allergen handling to ensure recipes are safe for your dietary restrictions
- Recipe variety with four different AI-generated styles to choose from
What Users Can Do
Get Recipe Variety & Choice
Every search generates four different recipe styles to choose from. Whether you prefer detailed instructions or quick overviews, you'll find a recipe that matches your cooking style.
Cook with British Standards
All recipes use British measurements (grams, ml, °C), proper UK ingredient names (courgette, aubergine), and reference ingredients available at UK supermarkets. No more converting cups to grams.
Stay Safe with Allergen Handling
Indicate your dietary restrictions and allergens when searching. The app handles all 14 major allergens as per Natasha's Law (including celery, gluten, crustaceans, eggs, fish, lupin, milk, molluscs, mustard, peanuts, sesame, soya, sulphites, and tree nuts), automatically ensuring recipes are safe and suitable for your specific needs.
Plan Your Entire Week
Beyond just finding recipes: organize your meals with a weekly calendar planner, automatically generate shopping lists that combine all your planned meals, and track pantry staples to avoid buying what you already have.
See It In Action
Watch how the app generates four different recipe styles in real-time
Recipe Variety - Choose Your Style
Every search generates four different styles: Balanced (comprehensive), Guided (step-by-step), Streamlined (efficient), and Essential (quick reference). Choose the one that matches your cooking preference. (Click image to enlarge)
Weekly Meal Planning Made Simple
Organize your entire week with the calendar view. Plan breakfast, lunch, and dinner, then automatically generate a shopping list for all your planned meals. (Click image to enlarge)
Why This Example Matters For Your Business
You might be thinking: "I don't need a recipe app—I need a booking system for my business" or "I need an inventory tracker" or "I need a customer portal."
That's exactly the point. This project demonstrates transferable skills:
- Building custom web applications from scratch that solve specific problems
- Creating user-friendly interfaces that people actually want to use
- Integrating complex services (payments, calendars, authentication, APIs)
- Thinking about the complete user journey, not just individual features
- Shipping finished, professional products that are ready to use
If I can build a recipe app with meal planning, shopping lists, and AI integration—I can build your booking system, inventory tracker, or customer portal. The industry changes, the skills don't.
⚠️ Important Notice
This is a portfolio demonstration. The app is fully functional, but recipes haven't been tested in real kitchens. Development is paused pending implementation of proper safety testing before any commercial launch.
Technical Implementation Details
For developers and agencies interested in the technical architecture, tools, and lessons learned
Why I Built This
I wanted to demonstrate my ability to architect and ship a complete SaaS application while solving a real problem I personally experienced: dinner decision fatigue. The multi-AI approach serves two purposes: it gives users variety (different recipe styles), and it showcases my ability to integrate multiple complex services.
The UK market focus came from research showing most recipe apps are US-centric with American measurements, ingredient names, and availability assumptions. British users represent an underserved market segment, and localizing for them demonstrates market research and product thinking skills.
I chose to build cost tracking from day one because understanding AI economics is critical for any production application. The analytics dashboard helps make data-driven decisions about which providers to use and when.
Project Overview
Platewise generates personalized recipes using artificial intelligence, specifically targeting UK home cooks. Users input available ingredients and dietary restrictions, then receive four different recipe styles in approximately 30 seconds from different AI providers, allowing them to compare and choose their preferred approach.
This project demonstrates my ability to architect and direct a full-stack SaaS application, integrate multiple AI providers with intelligent routing, implement proper authentication and data security, build comprehensive admin tools and analytics, and deploy a production-ready application.
My Role: Technical product builder using AI assistance. I made all architectural decisions (which AI providers, complexity scoring design, feature prioritization, tech stack choices) and configured all services, while using AI tools to assist with code implementation.
Tech Stack
Technical Architecture
Multi-AI Provider Integration
Integrated 4 AI providers (OpenAI, Anthropic, Google Gemini, XAI) with parallel processing and intelligent routing. Each provider generates unique recipe styles simultaneously, with error handling and fallback mechanisms.
Sophisticated AI Prompt Engineering
Custom prompts.ts configuration provides each AI with comprehensive UK food safety documentation, including all 14 Natasha's Law allergens with detailed guidance. Combines domain expertise with AI capabilities to ensure culturally-appropriate British recipes with proper allergen handling and safety prioritization.
Comprehensive Cost Tracking
Every AI request is logged with token counts, costs, and performance metrics. Admin dashboard shows total costs, average per recipe, and provider comparison analytics for data-driven optimization.
Complexity-Based Request Prioritization
Requests with allergens or complex dietary needs get higher priority scores, ensuring safety-critical requests use more capable AI models. Demonstrates intelligent routing logic and risk management.
Admin Review & Publishing Workflow
Admin users can review AI-generated recipes, edit SEO metadata and allergen tags, and manage publication. Uses Clerk user metadata for role-based access control with JWT session claims.
Admin Tools & Analytics
AI Cost Analytics Dashboard
Real-time tracking of AI usage costs, token consumption, and provider performance. Shows total costs, average cost per recipe, and identifies the most cost-effective providers for data-driven decision making. (Click image to enlarge)
Admin Recipe Review & Publishing Workflow
Comprehensive admin dashboard for reviewing AI-generated recipes, editing SEO metadata, managing allergen tags, and controlling publication to public recipe pages. Demonstrates role-based access control implementation. (Click image to enlarge)
What I Learned
Service Integration
Configured Clerk authentication with Supabase using the 2025 recommended integration method. Set up Google SSO, managed JWT session claims for admin access, and implemented Row-Level Security policies.
Database Evolution
Managed 31 database migrations showing how schemas evolve as requirements become clearer. Learned when to use JSONB for flexible data structures vs. normalized tables.
Cost Awareness
Built comprehensive usage tracking from day one. Logging tokens, costs, and performance metrics enables data-driven optimization and sustainability planning for SaaS economics.
Professional Judgment
Paused development when I realized AI-generated recipes require human testing before commercial use. Technical completion doesn't equal readiness for public deployment.
Project Status
Feature-complete MVP deployed for demonstration purposes. The application is fully functional and live in production, but not commercially available. Development is paused pending implementation of human testing protocols required before AI-generated recipes can be safely published.
Next steps if resuming: Implement human testing workflow for recipe validation, add chef review notes and approval system, verify allergen safety with qualified professionals.