On the MIRS, we encourage students to test their talent in real environments.Next, Elchin explains how the technical knowledge and resolution skills acquired enabled him to lead his team to victory:

"A group of strangers come together, form a team and end up winning the whole competition. That’s exactly what happened to us at the LeRobot Hackathon in Bilbao.

The idea quickly came about: to create an “AI Bartender”. We wanted to go beyond a simple robot and proposed a complete system: from voice to text, passing through an LLM (Gemini) to give it personality, and finally executing the action with the robot. To train it, we used imitation learning, collecting data on precise movements to serve ‘water’ (blue sand) and ‘coca-cola’ (red sand).

It was not an easy path. We had to reinstall the entire operating system in the middle of the hackathon and use cloud computing with Nvidia’s GPUs to train the ACT model in time, as our laptops were not enough.

During the final demonstration, with a large crowd observing, something unexpected happened. As we couldn’t access the table to change the cup because of the crowd, the robot served both drinks in the same container. What could have been a mistake turned into an improvised ‘cocktail function’ that was a hit with the judges.

From strangers to winners in a weekend. Seeing the name of our team in first place made 48 hours of effort worthwhile.'

If you want to watch the video of the demonstration and learn about the technical details of the code and the training, I recommend reading the full article:

Read the full story: ‘How we won the LeRobot Hackathon 2025’ at elchinaslanli.com