Fuel your Fitness

Inspiration

The challenge question itself: Design a tool to suggest healthy meals and find fresh food nearby. While there are various apps with individual features, this is an all-in-one solution for your dietary needs, specifically tailored to the fitness demographic.

What it does

The app prototype allows users to log in/register and provides three main sections: Profile, Eat In, and Dine Out. Profile Page: Uses the Mifflin-St Jeor Equation to calculate daily caloric needs based on age, weight, height, and activity level. This helps generate a personalized diet plan. Eat In: Users can add home-cooked dishes and track protein, carbs, and calories consumed. This feature runs on a Large Language Model (LLM) that analyzes ingredients and nutritional values. Dine Out: Displays restaurant dishes nearby that align with a user’s dietary needs. The system is trained on a dataset of New York restaurants, utilizing DeepSeek to classify dishes based on nutritional content.

How we built it

Frontend: Built using Next.js, allowing seamless integration with the backend and improving performance. Eat In: Integrates an LLM API that processes user-input meal descriptions to provide nutritional breakdowns. Dine Out: Uses a curated dataset of restaurant dishes, annotated with dietary attributes (e.g., low-carb, high-protein, vegan, vegetarian). DeepSeek-R1 is leveraged to analyze and classify menu items. Location Services: Google Maps API is integrated to find restaurants near the user and display them based on dietary preferences.

Challenges we ran into

Fine-tuning the LLM to accurately interpret meal descriptions and calculate nutrition. Training the DeepSeek model to correctly classify restaurant dishes according to dietary requirements. Ensuring smooth integration between different APIs (LLM, DeepSeek, Google Maps) while maintaining a seamless user experience.

Accomplishments that we're proud of

Successfully integrating AI-powered meal tracking and restaurant recommendations into a single platform. Training a dataset specifically for New York restaurants that enhances dietary classification accuracy. Implementing a robust user profile system that dynamically updates diet plans.

What we learned

The importance of accurate nutritional data when building AI-powered meal-tracking solutions. How to effectively train and utilize DeepSeek for dietary classification. Best practices for integrating multiple APIs into a single application while optimizing performance.

What's next for Fuel Your Fitness

Expanding the restaurant dataset to cover more cities. Enhancing the AI model’s ability to recognize more complex meals and recipes. Adding meal-planning features based on user preferences and goals. Implementing barcode scanning for quicker meal logging. Incorporating community-driven features, such as user-generated meal plans and restaurant reviews.