On the 15th edition of the Tech Up nights, we dove deep into pure data science to discover the algorithms MediaMarkt uses for stock optimization, and learned about Eurecat’s tools for summarizing audiovisual news content. Content created by techies and for techies, with a good dose of networking.

Carlos Nodar, Lead Product Owner at MediaMarkt & Aleix Sacrest, Data Scientist at MediaMarkt

MediaMarkt showed how Data Science is changing the way products are organised, using advanced models to work out how much stock to keep in store and choose the right mix of products. One important new idea is using personalised advice to work out the best amount of each product to have in each category and store. By using information about past sales, future demand and real-time trends, these models help to avoid having too many products and make sure that products are available in the places where they are needed most.

They are also using large language models (LLMs) to analyse web shop searches, which gives them valuable insights into what customers want. By finding patterns in how people search, the team can make sure that the right products are prioritised. This means they can work better and also make shopping better for customers by selling the right products at the right time.

Rafael Redondo Tejedor, AI and Computer Vision Researcher, and Head of the Image Group at Eurecat’s Multimedia Technologies Unit.

Eurecat is working on technology that uses AI to make short and clear summaries of long videos. This makes it easier to look for and use videos in different ways. By combining large language models (LLMs) with speech transcription, this technology extracts the most important moments in news, interviews and events. This makes the content easier to access and engage with.

The technology uses advanced speech-to-text models and is easy to use, as validated by media professionals. These smart tools can compress content really well, so you can summarise a 10-minute interview into just one minute. This helps to make the editing process more efficient, while keeping the quality of the content high.

However, “human supervision is still essential to ensure content quality”.