FoodGenie

Your Health, Engineered, Instantly. Stop Guessing, Start Eating Smart.

Comment

🥗 FoodGenie: The Personalized Nutrition Navigator

💡 Inspiration: The Consumer Problem

Today, going to the grocery store is an ordeal, an informational overload. Customers spend huge amounts of time agonizing over food labels, attempting to reconcile the facts with their own personal wellness goals and dietary restrictions. FoodGenie closes this gap by transforming obscure label data into immediate, personalized, and actionable health recommendations.

🎯 What It Does: AI-Powered Personalized Analysis

FoodGenie is a smart health assistant that provides instant, evidence-based recommendations tailored to the user's health profile.

  • Multimodal Input: The user simply uploads a single photo containing both the nutrition facts panel and the product barcode.

  • Intelligent Data Fusion: It uses Pyzbar for reliable barcode scanning and OCR for label text, then fuses this data with information from external nutrition APIs.

  • RAG-Enhanced Analysis: The extracted data is sent to an AWS Bedrock RAG agent (Grounding provided by FDA/USDA guidelines).

  • Personalized Verdict: The agent returns a detailed summary that directly checks the product against the user's health goals, allergies, and chronic conditions, providing a safety verdict and suggesting alternatives.

⚙️ How We Built It: A Serverless AWS Ecosystem

This application is a full-stack, serverless solution demonstrating integrating multiple AWS services:

Frontend: Python/Streamlit for a fast, intuitive UI.

Core AI Logic: A multi-step pipeline utilizing AWS Bedrock for the RAG-based, personalized analysis.

Data Persistence: MongoDB Atlas for saving and retrieving user history, allowing them to track past product analyses.

Backend & Orchestration: AWS API Gateway integrated with Lamba handles the complex multimodal data flow, API calls, and interaction with the AI.

Image Processing: Deployed with CV libraries (Pyzbar and Pillow) for robust barcode detection. Utilized Amazon Textract for OCR Recognition of nutrition facts.

🚧 Challenges

Challenge: Initial OCR efforts failed dramatically due to angled or poor-quality barcodes.

Solution: We pivoted to a Computer Vision/CV-based solution (Pyzbar) for barcode detection, making the entire system robust against real-world user photography. We successfully integrated a cohesive chain of AWS services (Bedrock, Lambda, MongoDB) to deliver a single, useful output.

🚀 What's Next For FoodGenie

The future holds much in store for tailored AI assistants:

Extended RAG: Can integrate a database of known food alternatives to provide better, geo-located product recommendations.

Real-Time CV: Implement advanced CV to better detect and localize nutritional text blocks, reducing reliance on barcode availability and improving data accuracy across diverse packaging formats.

E-commerce Integration: Use the personalized data to instantly filter and rate products within online grocery platforms

Health Customization: Can personalize recommendations further by integrating more detailed health information for users, and connecting with their digital health app.

Built With

Share this project:

Updates