Recipe-to-Cart Assistant
Pick a recipe and AI matches every ingredient to real products available on an online grocery store.
Try It Yourself
Search (swipe left or right) for a recipe, browse the ingredient list, and add everything to your cart in one tap. Recipe-to-Cart automatically matches every ingredient to real products on your grocery platform, so you go from recipe idea to ready-to-checkout cart without the back-and-forth.
The Problem
Finding Asian ingredients online is tedious. Recipe ingredients like 'scallions' don't match product names like 'Green Onion 1 bunch.' Users spent 30 minutes manually searching for each ingredient, leading to frustration and abandoned recipes.
What does this mean for a grocery platform?
Creating an open ecosystem that makes the entire internet's recipes shoppable.
External Platform Integration
Before: Users manually create recipes within Weee!
After: AI auto-matches ANY recipe from Instagram, AllRecipes, Reddit → instant shopping
Tap millions of existing recipes vs waiting for user-generated content
AI-Powered Matching
Before: Manual product tagging (time-consuming, inconsistent)
After: AI matching with explanations and alternatives
Faster creation, better quality, consistent experience
Smart Substitutions
Before: Out-of-stock products = abandoned recipes
After: AI suggests contextual replacements ("Use bok choy instead of napa cabbage")
Maintain purchase intent despite inventory gaps
What I Did
My Role
Product Manager + UX/UI Designer + Engineer
Owned every layer solo, from designing the user experience and making product decisions to building the full app with an AI ingredient matching engine. A deliberate challenge to prove I could take a product from concept to working app entirely on my own.
User Journey Flow
How the matching works under the hood
Pre-Filter
Narrow 2,000+ products to top 20 candidates
AI Match
GPT-4o-mini provides best match + alternatives
Smart Cache
Instant retrieval for repeated ingredients
App Screenshots
Clean, minimal recipe view that's easy to scan at a glance.
Add matched ingredients to cart one by one, or add the entire recipe.
AI recommends the best product match, with the option to search manually.
AI explains exactly why it's the right choice for the dish.
Results & Impact
Successfully deployed on Vercel with comprehensive documentation, ready to scale.
What I Learned
- •Building with AI gave me a new appreciation for my own work as a PM. I had written PRDs and led planning before, so I thought I understood why they mattered. But building this project myself made me feel it from the other side. When the tool interprets your intent and fills in the gaps on its own, a vague direction gets built out quickly in the wrong way.
- •Responsible AI design is also about earning trust. Building this while taking Human-AI Interaction at CMU meant learning and applying at the same time. It made clear how leveraging AI's capability and deliberately designing for trust have to go hand in hand, so that both the builder and the user can have real confidence in the product.
- •Working with AI made me a more curious product manager. Getting into the algorithm behind the matching experience revealed how much the structure of a solution shapes the quality of the user experience. That felt like a new layer of product thinking, and one worth keep building on.