Inspiration

  • Despite the availability of a diverse array of 63 cuisines, the average person typically sticks to just 3 to 4 preferences.
  • Language is another barrier in understanding the menu for many of foodie people.

What it does

DishDiscover uses cutting-edge Generative AI to revolutionize meal discovery. Our platform delivers personalized food recommendations tailored to your tastes, dietary preferences, and mood. Based on cuisine you like to try or picture of the restaurant menu you are visiting.

How we built it

  • Gemini-Pro and Gemini-Pro-Vision are The backbone of DishDiscover.
  • Firebase is used to handle all the backend processing. a. Firebase-Authentication is used to quickly login and authenticate users with google single sign on. b. Firebase extension for Multimodal Gemini API is used along with Firestore Database to generate the recommendations based on an engineered prompt. c. Firebase hosting is used to publicly host the WebApp enabling lean development and continuous deployment.
  • Figma is used for designing the user interface.
  • Flutter and Dart is used for frontend development.

Challenges we ran into

  • The biggest challenge we faced was to fine tuning the prompt. As we are providing the personalized recommendations, the user profile we are building is very crucial. We put in a lot time and efforts in prompt engineering to understand what data points we need to have in a user profile.
  • Another challenge we faced was with designing a user friendly application. We did lot of brainstorming and went through lot of iterations before finalizing our user interface.
  • Lastly hosting the application online and serving the public requests was another challenge as we had to minimize the infrastructure cost and the latency for the response.

Accomplishments that we're proud of

  • We are the only one providing exclusive recommendations from wide range of cuisines with just couple of clicks, allowing users to explore new dishes and satisfy their taste buds.
  • We are very proud to be pioneers in implementing a workflow allowing users to just click a picture of a restaurant menu to get instant personalized suggestions. This is very unique and shows the real power of latest innovations in Multimodal Generative AI.

What we learned

  • It is crucial to design a good generalized prompt to unlock the real potential of the Generative AI
  • User interface is as important as the backend for delivering a great customer experience.
  • New features like Google AI Studio and Firebase Gemini Extensions makes the application development very easy and quick.

What's next for DishDiscover

  • Enabling users to order the recommended food items by integrating DishDiscover with UberEats, DoorDash.
  • Providing options to get the recipes and groceries for cooking the recommended dishes.
  • Showing nearby places which serves the recommended dishes with great ambiance and dining experience.

Built With

  • dart
  • figma
  • firebase-authentication
  • firebase-extensions
  • firebase-hosting
  • firestore-database
  • flutter
  • gemini-pro
  • gemini-pro-vision
  • google-ai-studio
Share this project:

Updates