Eat & Care

Where clinical intelligence meets your meal—the only app that checks your food for safety and synergy which focuses on the unique food compatibility/safety engine

Comment

Inspiration

The inspiration for Eat & Care stems from a common gap in the digital health market: most apps tell you what to cook, but very few tell you why certain foods might be hurting or helping you based on your current biological state. I wanted to move beyond the simple calorie counter. I were inspired by the concept of Food as Medicine—the idea that during times of illness (like the flu or menstruation) or when managing allergies, the kitchen should be a place of healing, not a source of digestive stress or nutrient inhibition.

How to Built It

The project was built using a modern React + TypeScript stack, focusing on a "Clinical-Premium" aesthetic. The heart of the application is a custom integration with Gemini 3 Flash, which acts as our "Nutrition Intelligence" layer.

The build process followed three core phases:

  • The Logic Layer: I developed a filtering system that doesn't just look for ingredient names, but understands biochemical relationships. For example, the system uses logical arrays to cross-reference user inventory against a contraindication matrix.

  • Multimodal Integration: I implemented a "Scan Pantry" feature using Gemini’s vision capabilities, allowing the app to translate a simple kitchen photo into a structured JSON list of ingredients.

  • Persistence & Localization: I leveraged the browser's localStorage to create a "no-login-required" experience that still feels personalized and permanent.

What I Learned

Throughout this journey, I learned the true potential of Multimodal AI. Moving from text-based prompts to image-based ingredient recognition taught me how to handle unstructured data effectively. I also gained deep insights into Internationalization (i18n), learning that true translation isn't just about changing words—it's about re-prompting the AI to provide culturally and linguistically relevant health advice in real-time.

Challenges I Faced

Visual Accuracy: Early versions generated generic food images. I overcame this by implementing the Gemini 2.5 Flash Image model to generate bespoke, professional-grade photography that actually reflects the specific recipe generated.

Built With

Share this project:

Updates