Food Haven

(3-Layer Smart Storage)
Time-Sync & Spoilage Alert
Multi-modal AI Recognition)
Smart Recipes & Harvest Guide

Inspiration

As an international student in Denmark with zero cooking experience, I struggled with food management. I often faced moldy vegetables from wrong storage and even witnessed a friend being hospitalized after mistaking toxic daffodils for garlic sprouts. These life-threatening mistakes and the waste of resources inspired me to create Food Haven—a digital "expert sense" for young people living alone.

What it does

Food Haven is a digital sanctuary for ingredients. Powered by Gemini 1.5 Pro, it offers:

Smart Identification: Instantly recognizes single or multiple ingredients from photos and suggests optimal storage (Pantry, Fridge, or Freezer).

Real-time Tracking: A time-slider feature syncs with real-world time to predict food status and send alerts before spoilage.

Harvest Guide: An encyclopedia providing fun facts and preservation methods for new ingredients.

AI Recipe Generator: Recommends detailed recipes based solely on what is currently in your "Haven".

How we built it

Natural Language Development: As a team of designers with no coding background, we built this entire project by using natural language to communicate our vision to Gemini 3. This allowed us to transform complex design concepts into functional app logic without writing a single line of manual code.

Gemini-Powered Core: We used Google AI Studio to refine our prompts, enabling the model to handle all backend reasoning—from identifying various ingredients to predicting complex spoilage timelines.

Multimodal Vision Logic: We leveraged the advanced multimodal capabilities of Gemini to process high-density visual data. This makes it possible for the app to categorize dozens of different ingredients simultaneously with just one click.

Designer-to-Demo Prototyping: The interactive 3-layer storage system and the time-sync engine were conceptualized in our design tools and brought to life through AI-assisted development, moving directly from a vision to a working demo.

Challenges we ran into

Recognition Precision: While the AI is powerful, ensuring 100% accuracy in ingredient recognition remains a hurdle. Real-world factors like tricky lighting or overlapping items can sometimes affect the precision of the scan.

Backend Constraints: Because of the tight competition timeline, our backend data persistence system isn't fully finished yet. This means the data is currently processed in real-time sessions rather than being stored long-term.

API Rate Limits: Since we are running the demo on the Gemini free tier in AI Studio, we occasionally hit rate limits when performing multiple high-density image scans in a row.

Accomplishments that we're proud of

We are incredibly proud to have built this project independently as designers with no prior coding background, proving that Gemini allows us to bridge the gap between creative vision and reality through natural language. We successfully delivered the core requirements: the ingredient recognition is highly accurate, and the storage logic for the pantry, fridge, and freezer works seamlessly.

We also put a lot of heart into the user experience, ensuring the UI is visually polished and the animations—especially the real-time time-sync slider—are smooth and intuitive. It’s rewarding to see the AI handle complex photos of multiple ingredients to enable "one-click" categorization, perfectly linking our inventory system to smart recipe recommendations to help users reduce waste.

What we learned

As a team of designers with no prior coding experience, our biggest takeaway was learning how to independently build a functional application from scratch by leveraging AI-assisted development tools. We discovered that while the AI engine is powerful, a clear information architecture and an intuitive UI are just as important as the underlying technology when it comes to helping users navigate something as critical as food safety. This journey taught us how to use natural language to communicate complex logic to Gemini, effectively turning our design intuition into a working prototype that solves real-world problems