Culinary curator
This proof of concept began as a response to my own decision fatigue as a frequent home cook overwhelmed by endless food content. My goal was to move from scattered inspiration to calm, confident execution without undermining agency.
To solve this, I vibe-coded a working prototype of a multimodal AI companion that transforms chaotic, entertainment-optimized recipes into clear, personalized, repeatable flows. I led the end-to-end AI experience strategy and trust model. Built with React, OpenAI APIs, and multimodal parsing, and published via Figma Make; the system extracts and structures recipes from videos, images, and links into guided, low-friction cooking flows.
-
Extracts ingredients and steps from social content
Clearly labels exact vs AI-inferred information
Adapts instructions to diet, time, skill level, and energy
Provides step-by-step, hands-free guidance with timers and cues
Suggests smart substitutions using pantry awareness
-
Users scan pantry items or paste a recipe link, image, or video they already trust.
The system transforms scattered inspiration into a clear, structured recipe with visible confidence markers.
Step-by-step, hands-free guidance reduces cognitive load through pacing, timers, and cues.
Smart substitutions and adaptations adjust for diet, time, and skill without removing agency.
Post-cook feedback quietly improves future recommendations, reinforcing ease over effort.
-
Preserve culturally rich recipes without flattening context or erasing authorship.
Clearly separate source truth from AI inference to maintain trust.
Make AI intent and confidence visible through transparent design.
Augment human expertise rather than replace it.
Promote wellness, sustainability, and equitable knowledge sharing.
Eventually support multiple languages, abilities, and learning styles with accessible, multimodal guidance.
My role and key decisions
I shaped the overall product direction and AI experience architecture, establishing the principles for trust, transparency, and multimodal interaction from concept to working prototype. I prioritized explainability over automation, clearly separating source truth from AI inference and avoiding fully generative recipes to reduce hallucination risk. The system was designed as a unified, scalable AI infrastructure focused on long-term trust, habit formation, and measurable user value.
Projected impact
Culinary Curator is designed to reduce everyday cooking friction and food waste (est. 5–10%) by turning scattered inspiration into structured execution, potentially lowering impulse takeout and saving users ~$15–$30 per month. Cooking frequency could stabilize at 2–3 meals per week, signaling durable habit formation. By prioritizing transparent AI over novelty, the product builds long-term trust, strengthens retention, and supports scalable subscription and wellness revenue.