Smart Menu

• 

  Motivation

• 

  Our Solution

• 

  Our Product

• 

  Tech & Algorithm

Inspiration

Nowadays, restaurant industry offers a wide range of dishes. People cannot find the right dishes without trial and error.

What it does

Our application offers customized restaurant menus for individual restaurant patrons to maximally enrich their experience by recommending dishes and drinks that each of them is likely to enjoy most.

How I built it

We use Ionic for the front end/mobile-app development, and Django RESTful for the backend. Our recommender engine is based on a hybrid approach between collaborative filtering and dish-based recommendation with uses of mechanisms to garner patrons’ implicit feedback from application usages to guide the recommender engine. We bootstrap our recommender with data from Point-of-Sale machines from restaurants.

Challenges I ran into

Varieties in different recipes, tastes, cooking styles, and presentations make it very hard to profile a dish.

Accomplishments that I'm proud of

We have built a prototype that has (almost) all functionalities of the entire system using real data form PoS machines in just two-and-a-half days. We also formulated some new ideas about how we might better profile dishes.

What I learned

1. It is quite hard to profile dishes efficiently.
2. Using and getting implicit feedback well is quite a challenge that requires thinking about the whole system holistically. ## What's next for Smart Menu
3. Improve the recommender engine in the prototype.
4. Market/deploy it in restaurants to garner more data.