UniRec

Unified neural recommender system combining content and product recommendations using StarSpace for B2C and B2B clients.

Overview

UniRec is a unified recommender system combining content and product recommendations for B2C and B2B clients. Utilizing a neural StarSpace model, it continuously learns from user behavior to provide personalized, real-time recommendations.

Architecture

  • Neural StarSpace model for embedding-based recommendations
  • Consolidated multiple legacy systems into a single platform
  • Scalable, easy-to-deploy solution within the Optimizely ecosystem

Tech Stack

Python TensorFlow StarSpace GKE BigQuery

Impact

  • Boosted engagement by 12% and revenue by 8% for B2C/B2B clients
  • Processing 50M+ daily recommendations
  • Notable clients: Australian Red Cross, Outrigger, NewEra, The Royal Mint