An AI‑powered platform for fashion discovery

As an organisation, we are focused on innovation and strive to provide our users with the best experience possible. With 15 million fashion products, GLAMI is more than a catalog - we are a discovery platform! The personalised discovery we enable at scale helps shoppers find pieces they love.

Innovation is at the core of what we do, and we continually work to deliver the best experience for our users. GLAMI is more than a catalog of 15 million fashion items- we are in the discovery business. We help make it easier to discover fashion pieces that shoppers love through personalized discovery at scale.

"E‑commerce and fashion are getting a makeover. First of all, we believe that one-size-fits-all model in the world of online fashion commerce is now outdated and that technology makes it cheap and easy to build a personal offer for every unique user. Secondly, GLAMI popularizes modern Machine Learning by bridging the gap between research and application, between complex models and easy‑to‑use features."

Antonín Hoskovec

Head of AI

Recommender system

We know what users may like

Recommender system helps solve the problem of browsing through the overwhelming amount of items in fashion. Similar to other projects that display search results, we know that users tend to pay most attention to the first handful of products shown. In some categories, that's only tens of products out of tens of thousands. Without personalization the same list of items would be shown to every user resulting in most of the items being hidden further down the list or on the next pages, never being seen. Showing relevant products is therefore not only beneficial to users but also to retailers. Our system generates about 60 million recommendations every month ensuring a tailored experience for every visitor.
Read more
Deep Item Categorization

Automatization and understanding

Deep Item Categorization automatically assigns the category and all applicable filters to the majority of products in our catalog. On average 10% of items that are on offer change every month. Keeping up with this momentum manually is impossible, thus we employ state‑of‑the‑art neural network architectures to take over the workload.
Read more


Recommendations per month


Recommendation models trained monthly


Tags assigned by AI per month


Similarity requests daily


Helping find similar pieces

Similarity is at the core of our machine learning. To find similar products, the model processes each product into a set of features that describe it. The same set of features is then used in several projects connected to Similarity such as the grouping of identical products. The system can handle around 3 million requests per day searching the most similar products in 15 million items catalogue.
Read more
cloud native

Cloud native AI infrastructure

We have built our entire AI infrastructure on the cloud. We are deploying and scaling all the projects automatically in large clusters (tens/hundreds of computers) of specialised hardware. The team has worked hard on a system that automatically sets up the required infrastructure for us and then deploys the models and code into it. In this way, we are able to continuously deliver our most recent codebase to the users.