dAIningRIT

when dining hall meets AI
Choose preferred cuisine
Choose ingredients
See your own recipe! (you can like them too)
See others' recipe

Inspiration

The dissatisfaction with college dining halls led us to create dAIningRIT, an AI-powered mobile web app that aim to make college dining more enjoyable. By combining already made ingredients in the dining hall, the student can make virtually any dish from anywhere they want, increasing enjoyment and reducing food waste!

What it does

Every day, dAIningRIT will automatically retrieve menu data from Gracie's, the largest dining hall in RIT, and generate personalized food recommendations for college students. The app takes into consideration what type of cuisine the student wants to eat, along with their preferred list of ingredients. It will then generate a recipe containing a list of ingredients and the station where they can be found and step-by-step instructions to create their desired food.

For example, if Gracie's only offers American food tonight, and the student is hungry for some Vietnamese food, they can come on dAIningRIT, pick "Vietnamese" as their cuisine, and the app will generate a fresh stir-fry Pho recipe using the ingredients at the dining hall!

The app also lets students see other people's generated recipes, so if they do not have anything in mind, they can get inspiration from others'. The app allows the students to vote for any good recipes they found, by which the app will sort the ranking. Dining hall administrators can make use of this information to guide their food selection for subsequent day.

How we built it

For recipe generation, we make use of GPT-4 (for instruction) and DALL-E 3 (for dish image generation). We host our backend on Google Cloud Functions with Python, our database on Firebase Firestore, our image storage on Firebase Storage, and deploy on Firebase Hosting. For the frontend, we use React and node.js. We use domain name from GoDaddy

Challenges we ran into

Since our team does not have much have experience in UI/UX, we have trouble coming up with a presentable web design.

Accomplishments

We successfully integrated GPT-4 and DALL-E-3 APIs to generate recipes after iterations and tweaking of prompts. This is our first time using LLM models, so we were somewhat confused, but everything worked perfectly in the end.

What we learned

We learned a lot about the prompting techniques to maximize the use of GPT-4 and DALL-E-3, which we will be able to use in the future. We have also gotten more experienced in deploying on Firebase as well as doing frontend engineering work.

What's next for dAIningRIT

Our journey doesn't stop here. We are committed to ongoing improvements for dAIningRIT, with the goal of expanding its reach beyond RIT. Our vision is to make it a versatile solution that can be applied to various universities, providing personalized food suggestions to enhance the dining experience for students across different campuses. We hope to be able to also reduce dining hall's food waste by being able to customize the food selection to a larger body of people.