Inspiration We’ve all faced the frustration of jumping between multiple apps — one for groceries, another for recipes, and yet another for calorie tracking. This fragmented experience not only wastes time but also leads to poor dietary consistency. MealCart was inspired by the idea of creating a unified experience — where meal planning, grocery shopping, and health tracking coexist seamlessly under one roof.
What It Does MealCart bridges the gap between meal planning, grocery shopping, and health management. Users can: Create a personalized profile with dietary preferences, allergies, cuisine choices, and health goals. Converse naturally with an AI shopping assistant to get meal plans, recipes, and nutritional insights. Automatically generate grocery lists or add ingredients directly to their cart — no need to switch apps. Log meals to track nutritional intake and ensure alignment with health goals. The system intelligently recommends recipes that fit the user’s diet, budget, and skill level, while also leveraging the store’s product catalog for real-time availability.
How We Built It We built MealCart using a modern AI-driven, cloud-native architecture powered by AWS: Frontend: React app hosted via AWS CloudFront & AWS S3 for a smooth, responsive user experience. Backend: FastAPI app deployed using AWS Elastic Beanstalk for scalable and managed infrastructure. AI Layer: AWS Bedrock AgentCore and Agent Runtime to deploy and orchestrate AI agents. AWS Nova for conversational intelligence and natural language understanding. AWS Strands for data processing and retrieval-augmented workflows. Data Layer: Amazon DynamoDB for user profiles, preferences, and meal logs. Integration: The AI agent retrieves recipes, generates personalized plans, and integrates grocery data into a unified experience — all through an agentic orchestration pipeline. Challenges We Ran Into Designing a unified schema to connect user health data, meal preferences, and product catalogs seamlessly. Managing agent prompt engineering and orchestration across Bedrock AgentCore and runtime environments. Optimizing response times while ensuring data privacy and consistency in AI recommendations. Accomplishments We’re Proud Of Built a fully functional, end-to-end AI shopping assistant that connects meal planning with grocery shopping. Successfully integrated Bedrock AgentCore with multiple AWS services for contextual reasoning. Enabled a health-conscious, budget-aware AI capable of making intelligent, personalized food choices. Achieved real-time, conversational grocery cart automation from recipe generation to checkout. What We Learned Deep insights into agentic AI design — how to chain tools, APIs, and models effectively. Practical deployment knowledge of AWS Bedrock, Elastic Beanstalk, and DynamoDB in a production-style setup. The importance of contextual grounding in LLM-based systems to prevent irrelevant or inaccurate outputs. The value of designing user-centric conversational flows that feel natural and intuitive. What’s Next Expand to include fitness wearables integration for real-time health tracking. Add voice assistant capability for hands-free meal planning.
Log in or sign up for Devpost to join the conversation.