RamsayAI

Inspiration

The idea came from a common yet relatable moment: staring at a packed fridge or pantry and still thinking, "There's nothing to eat." We wanted to eliminate this everyday problem using AI. What if you could simply snap a picture of your kitchen inventory, and instantly know what meals you could make—based on what's there and what your body needs?

What We Built

Our project is a smart, AI-powered meal planner. Users can:

Take a photo of their pantry or fridge.
Let the AI detect ingredients and count what's available.
Generate recipe ideas based on their current inventory.
Specify calorie and macronutrient targets (like protein or carbs).
Get full recipes with instructions, cook time, and ingredient lists.
Even receive a full weekly meal prep plan—3 meals a day—based on what's on hand.

How We Built It

Frontend: Built with React, Vite, and Tailwind
Backend: Node.js
AI Integration: Used OpenAI's APIs to:
- Analyze food images.
- Generate meal suggestions and nutrition-focused recipes.
- Construct a full meal plan for the week.
Database: MongoDB stores ingredient data, user preferences, and meal history.

What We Learned

Working with vision models in real-world conditions (like dark or cluttered fridges).
How to structure prompts for clean, formatted AI outputs.
Building a seamless user experience for image-to-recipe generation.
Connecting multiple components (frontend, backend, AI, DB) into a cohesive system.

Challenges We Faced

One major challenge was dealing with poor lighting in user photos. Dim lighting and reflections made ingredient recognition unreliable. We tackled this by preprocessing the image to improve brightness and contrast before sending it to the AI.

Another issue was AI output consistency. We had to carefully craft prompts to ensure the AI returned cleanly structured recipes every time—complete with ingredients, steps, and macro info.