Case Study
An AI kitchen agent that sees what's in your fridge and tells you what to cook. Point your phone at the ingredients, get personalized recipes matched to your dietary preferences and time. Built end-to-end as a consumer AI product.
A consumer AI application available on iOS and web. You open the app, point your phone camera at your fridge or pantry, and Claude Vision identifies every ingredient it can see. Then it generates recipes matched to what you have, your dietary preferences, and how much time you want to spend cooking.
Each recipe comes with step-by-step cooking instructions, built-in timers, and substitution suggestions for missing ingredients. It's not a recipe database you search. It's an AI agent that understands what you have and what you want, and creates something specific to your kitchen right now.
Cross-platform mobile framework. Single codebase compiles to iOS and web. Camera access for ingredient scanning, native performance for the cooking timer interface, offline recipe caching.
Multimodal AI for ingredient identification. Processes camera frames, identifies individual items, estimates quantities, and handles messy real-world conditions: half-open containers, overlapping items, poor lighting.
Recipe generation engine. Takes identified ingredients, user preferences (dietary restrictions, cuisine type, cooking time, skill level), and generates complete recipes with nutritional estimates and substitution options.
Backend-as-a-service handling auth, user profiles, preference storage, recipe history, and favorite management. Row-level security for user data isolation. Edge functions for API orchestration.
Open the camera. Point at the fridge, pantry, or countertop. Claude Vision identifies ingredients in real-time: "chicken breast, bell peppers, garlic, soy sauce, rice."
Set your constraints. Dietary restrictions (keto, vegan, halal). Time available (15 min, 30 min, 1 hour). Cuisine preference. Skill level. Servings needed.
AI generates 3 recipe options ranked by match quality. Each shows cook time, difficulty, calories, and how many of your available ingredients it uses. Pick one.
Step-by-step instructions with built-in timers. "Sear chicken for 4 minutes per side" with a tap-to-start timer. Substitution suggestions if you're missing something.
I wanted to ship a consumer AI product end-to-end. Not a wrapper around an API call, but a real product: product spec, UX flow, technical architecture, AI integration, cross-platform deployment, and user data management.
Two platforms. Not a demo. A production-grade consumer app built with the same rigor you'd expect from a funded startup, by one person and an AI agent in a terminal.