Intelligent book recommendation service

Our partner needed a book recommendation system to improve customer experience and increase sales. The system had to make recommendations based on users’ browsing and purchasing behavior, but also had to offer relevant suggestions to first-time visitors. We developed two models:

1. An implicit feedback model that tracks user interactions, like page visits and purchases, to score books.
2. An explicit feedback model that analyzes previous user ratings to predict user preferences and adjust book recommendations.

Both models use SVD-like collaborative filtering to spot use patterns and make predictions about a user’s preferences. To feed the models and make the recommendation service reliable and responsive, we built a Scala-based app with the Play Framework and Akka Actors.

We then designed the recommendation engine to be a standalone service, so it deploys easily and retrains itself without disrupting the active flow of recommendations. As a final step, we deployed the service in a secure Kubernetes cluster for stability and efficient resource management during day-to-day operations and retraining.

Project outcomes

1. An accurate, scalable recommendation system that learns preferences in real time and runs uninterrupted.
2. Better customer retention and satisfaction with more relevant book recommendations.
3. Increased sales as recommendations align more closely with personal taste.