All projects

Project 04, Machine learning and full-stack

BiteKaro.

A self-ordering cafe kiosk that suggests what you might want next.

Capstone project score99/100for a kiosk with a hybrid menu recommender
Timeline
Jul 2025 to Apr 2026
Type
Capstone Project I and II
Score
99/100
Parts
Kiosk, admin and recommender

01, The idea

A kiosk that knows what goes well together.

BiteKaro is a self-ordering kiosk for a cafe. Customers browse the menu, pay and track their order on the screen, while the cafe manages everything from an admin dashboard.

We wanted the kiosk to do what a good waiter does, which is to suggest something that fits the order and the time of day.

02, What we built

Everything a small cafe needs on one screen.

Ordering kiosk

Menu, cart, Razorpay payments and loyalty features, built for a touch screen.

Admin dashboard

The cafe manages the menu, orders, customers and analytics in one place.

Live order updates

Order status updates arrive in real time through Socket.io.

03, How the recommender works

Three methods, blended into one suggestion.

  1. Bought together

    Apriori association rules find items that are often ordered together.

  2. People like you

    SVD collaborative filtering learns taste patterns from past orders.

  3. Right now

    A time-of-day popularity baseline brings up what sells best at this hour.

  4. Tested

    Training and evaluation scripts check the recommender against held-out orders.

04, Notes

What it was trained on.

  • Public data. The recommender was trained on a public bakery transactions dataset, remapped to an Indian cafe menu.
  • Capstone score of 99/100. Evaluated as Capstone Project I and II across Semesters 5 and 6 at GLS University.

Built with

  • Python
  • FastAPI
  • scikit-learn
  • Node.js
  • Express
  • Socket.io
  • PostgreSQL
  • Razorpay
  • Tailwind CSS