Study Bites

Discover meals near you that fit your budget, dietary preferences, and lifestyle in a few clicks. Our platform helps busy students find restaurants with tailored options and quick links.

Comment

Inspiration

Study Bites was born from the need to simplify meal decisions for busy college students. We noticed that during high-stress periods like finals, students often skipped meals or settled for options that didn't fit their dietary needs, cultural preferences, or budgets. We wanted to create something that not only reduces the stress of finding where to eat, but also respects students' diverse preferences to promote healthier food choices without the hassle.

What it does

Our project is a restaurant discovery website designed for students. It presents a curated set of food cards at a time—intelligently filtered based on users' unique location and needs, whether it's dietary, accessibility features, tight budgets, or family-friendly environments. Powered by Qloo's Taste AI API and Hugging Face Image LLM, the platform matches students with personalized restaurant recommendations, complete with ratings, distance, and actionable buttons to call, get directions, or visit the restaurant's website.

How we built it

We built Study Bites as a responsive web application using JavaScript, HTML/CSS, and Python Flask, featuring a food selection grid that is optimized for both mobile and desktop devices. It also features real-time geolocation and reverse geocoding to detect users' locations.

On the backend, we used an Nginx proxy and a Gunicorn production server to handle API calls and data processing pipelines. The project integrates Qloo's Taste AI API using tag filtering for selected preferences. To ensure image quality, we optimized displayed data with a custom CLIP-based ML model offered by Hugging Face LLM that classifies and validates restaurant images using a confidence score. As LLM processing takes some time, to enhance user experience, we also implemented concurrency and a thread-safe model in the background while serving web pages immediately right after results on the current page are available.

Challenges we ran into

  • API response optimization and efficient implementation to filter through food options
  • Ensuring high image quality of the restaurants and implementing CLIP with optimized model loading

Accomplishments that we're proud of

  • Successfully integrated Qloo's Taste AI API and CLIP ML model of Hugging Face LLM
  • Implemented a background image filtering system
  • Accomplished a robust error handling mechanism and responsive design
  • Enhanced Python Flask backend with multi-threading and response optimization

What we learned

  • Tag systems and insights of Qloo's API
  • A complete cycle of website development, from design, implementation, debugging, enhancement, to production deployment
  • Implement machine learning for real-time uses

What's next for Study Bites

  • Campus integration for campus meal options
  • Add group decision features where friends can collectively select preferences
  • Add features to predict meal preferences based on the time of the day and usage patterns

Built With

Share this project:

Updates