hoomeals

a web app that facilitates meal planning for UVA students.

Comment

Inspiration

As current college freshmen, we've seen firsthand how many college students want to start getting into fitness and care about their health - however, one of their greatest obstacles tends to be a dining hall. Most modern nutrition advice suggests careful calorie and macro tracking, but dining halls currently don't offer easy tools to do so.

What it does

HooMeals aims to solve this by providing both a convenient daily display for dining hall menus, an inbuilt calorie tracker, and features for figuring out how many calories and protein you should get per day. It also includes AI-powered recommendations to help students make better meal choices.

How we built it

Languages: HTML, CSS, JavaScript, TypeScript, Python Frontend: React.js, TailwindCSS, RECharts Backend: Node.js, Express.js, Beautiful Soup 4, Selenium, Bcrypt, MongoDB Machine Learning: Generative Embeddings from HuggingFace Tools: MongoDB Atlas, Vercel, Virginia Meals Website (https://virginia.campusdish.com)

Challenges we ran into

Initially, we ran into issues with our web scraper, which pulls data from the Virginia Campusdish. We were intially using Selenium for this, but due to the slow performance that the scraper was yielding, we decided to use Beautiful Soup to aide the scraping process, which optimized speeds. We also had some difficulty hosting the web scraper in Vercel. However, through this we learned more about Vercel deployment as well as the technologies mentioned above.

Accomplishments that we're proud of

This was our first time actually implementing a UI Workflow from Figma, which allowed us to work quicker and make intuitive and clean interfaces. We are especially proud of our menu and tracker pages, as the functionality of those pages took a while to finish up with React.js.

What we learned

We learned how to use and optimize web scraping tools in a deeper level using Beautiful Soup and Selenium. We also gained experience in database and API optimization in order to make our app efficient without waisting resources. Finally, we gained basic but practical knowledge in using pre-trained LLMs for larger use cases.

What's next for hoomeals

While we have a basic recommendation system that retrieves meals similar to the user's profile based on embedding similarity, there is still much room for improvement. We intend to improve the recommendation system to generate more accurate recommendations, using Mistral-AI to refine the process.

Since our website scraper is compatible with Campusdish, a website that is used by hundreds of unversities across the United States, the app has a potential to expand to multiple campuses, and we plan to leverage this after the improvements mentioned above.

Built With

Share this project:

Updates