There are a lot of apps for recommending meals at restaurants, but we eat at home most of the time and still need help coming up with feasible ideas that conform to our preferences.

What it does

It has 3 parts: 1) It updates a list of items in your cabinets by scanning multiple barcodes at the same time, with an object recognition approach in beta. 2) Like Pandora, it creates a profile of your preferences by asking for thumbs up/down on an intelligent sampling of different recipes, then applies machine learning clustering and similarity analysis to suggest new options. 3) Provides this functionality on an app, with a partner website written by first-time coders (who learned html for the first time this weekend!).

How We built it

We're running a node.js server with a mongo database that helps us combine Android functionality with Anaconda-supported data manipulation and machine learning through Python. Our website employs a simple html and css match.

Challenges We ran into

Getting useful data from recipe api's and cleaning it to provide meaningful recommendations. Learning MongoDB, and making the whole ecosystem work.

Accomplishments that We're proud of

Making various languages and diverse functionalities all play together nicely. And learning whole new languages, and even a new skillset (coding!).

What I learned

Please see above. :)

What's next for Chefit

Object recognition and refined models for more meaningful recommendations.

Share this project: