Sales Push Notifications Recommendation System

Choose is a mobile e‑commerce application that sends a daily sale to each user via push notification. The existing recommendation system relied on a simple popularity‑based algorithm, sending the most clicked sale from the "Upcoming sales" section to all users. This approach lacked personalization and failed to deliver tailored experiences, highlighting the need for a more sophisticated system to improve user engagement and open rates.

The main challenge was to build a scalable recommendation system capable of handling millions of users and thousands of sales while delivering personalized recommendations. In addition, we needed to address the cold start problem for new users and new sales, and strike a balance between exploitation (recommending items users are most likely to engage with) and exploration (introducing new and diverse sales to users).

I designed and implemented a hybrid recommendation system combining collaborative filtering, content‑based filtering, and deep learning approaches. The system used a neural network with a shared‑bottom architecture to simultaneously predict whether a user would open the push notification and whether they would click on the sale within the first 24 hours on the homepage.