Back to projects

Retail AI // recommender systems // scale

Retail recommendation system for 12 million customers

Large-scale retail recommendation project inspired by Google-style Two-Tower retrieval architectures, designed for millions of customers.

The project

The project focused on retail recommendations at scale: representing customers and products in compatible vector spaces, then using those representations for personalized retrieval and ranking.

Technical angle

The hard part is not only the model. You need clean transaction history, freshness, popularity-bias controls, temporal windows, usable features and production integration.

What it proves

It shows the ability to move from data science to business impact at a scale of millions of customers.