As part of the Hespresso theme magazines, the HES-SO Valais Wallis has written the following article on Calixte Mayoraz, a research and development engineer at Eversys SA and a specialist in machine learning. He studied computer science at the HES-SO Valais Wallis, and completed an internship with the computer science institute at the Swiss Digital Center in Sierre. He then completed a Master’s degree in Artificial Intelligence at UniDistance in collaboration with Idiap.
For many of us, AI is a kind of Moscow eye used by geeks to rule the world. We admire the machine’s potential, just as much as it scares us. Calixte, on the other hand, has been caught up in it from an early age. At the very time we were learning to pronounce “Kevin is in the Kitchen”, Calixte was working in Silicon Valley alongside his dad, a senior manager at Google. Perfectly bilingual, the native of Sédun completed his high school education in California before returning to his roots in Valais at the age of 18.
Whether it’s a question of genes or not, Calixte admits that he followed in his father’s footsteps by naturally turning to IT at the Haute Ecole de Gestion. His first work placement was with the Computer Science Institute, followed by a Bachelor’s degree and three years back at the institute as a research assistant. “With Professor Dominique Genoud, we worked on data and machine learning. It was very exciting. But deep down I felt that we were just scratching the surface of it all. To go further, Calixte did a Master’s degree in Artificial Intelligence at UniDistance and wrote a practical dissertation for Eversys SA, which still employs him today.
Eversys SA manufactures super-automatic coffee machines for professional∙le∙s. These machines are designed, tested and assembled in Sierre. In a nod to the American dream, last year the company won a major contract with Starbucks to produce cold coffees. Calixte’s job is to carry out predictive maintenance. In other words, to anticipate breakdowns: “As soon as someone interacts with one of our machines, as soon as a coffee is brewed somewhere in the world, it sends us a whole series of data such as the water temperature, the boiler pressure, the opening of the grinder, the rinsing, and so on. Using this information, we try to understand the breakdowns and adapt the settings. Note that our biggest machines produce between 2,000 and 3,000 coffees a day!”. Among the 300 employees∙e∙s, Calixte is the only AI specialist who knows how to handle statistics and all the finer points of IT. When he can’t find a solution to a given problem, ChatGPT becomes a good ally. “I contribute to a chat thread. It allows me to develop certain ideas and keep track of the proposals that the AI submits to me, which are usually interesting”. The idea of using ChatGPT to help Eversys’ technicians∙ne∙s with their tasks is also gaining ground…
A computer scientist and private DJ, Calixte loves everything to do with computers. Recently, he has found a new challenge: “I’m trying to do some artistic stuff where I deliberately use an AI that is poorly trained, so a blank neuron base, to do some collaborative drawing. A bit like an exquisite corpse, I start to draw something and the AI continues my drawing. The experiment serves to show how much the AI needs to be trained to recognise what I’ve sketched, so that the whole thing is coherent. It’s fun and I like to share these ‘aesthetic discoveries’ with my artist friends.
The limits of the machine
For Calixte, AI is an incredible tool, but one to be wary of. To the question “Will AI replace us all one day?”, his answer is nuanced: “It is capable of replacing us, but not always by doing a good job. If you don’t go back and check what it does, if you trust it blindly, you risk ending up with some nasty surprises.