“Lord of the recommendation models, Two Tower for All(egro)”
Building effective recommendation systems for a massive e-commerce platform like Allegro.com—with over 20 million active buyers and a highly dynamic catalog —presents formidable challenges. How do you deliver diverse, high-relevance suggestions across dozens of placements while keeping maintenance low? We present a unified, content-based recommendation system built on the Two Tower retrieval framework. We demonstrate how this single architecture is adapted to serve three distinct user intents: similarity search, complementary suggestions, and inspirational content discovery. Our solution handles dozens of thousands requests per second at production scale. Insights from two years of continuous A/B testing confirm statistically significant gains in business metrics, proving that flexibility and scalability are key for product development at Allegro scale.
Bio
Aleksandra is a Machine Learning Manager at Allegro, leading a team of Research Engineers. Her team’s role is to be the Indiana Joneses of e-commerce, digging into data to uncover the best ways to connect users with what they love, developing dedicated ML models for retrieval and ranking in recommendation systems. She traded her academic research career (yup, she’s one of those crazy ones with PhD in computer science), but she still maintains a passion for growth and learning new things. Whether it’s cracking a tough problem at work, hitting a tight forehand on the tennis court, or attempting to teach her Basset Hound, Rysio, that “sit” isn’t just a suggestion.