News
Pierre-Louis Bescond (SKEMA 2004): A pioneer at the helm of industrial AI
A computer science enthusiast since childhood, Pierre-Louis Bescond saw AI and machine learning as a new frontier for revealing previously elusive correlations. In charge of a team of data scientists and engineers at Roquette, he doesn't just manage - he continues to code, keeping a direct connection with the most cutting-edge projects and technologies. Convinced of the potential and limits of AI, he also warns against the risk of "intellectual inbreeding" of models. Under his leadership, Roquette is exploring new horizons, integrating AI in fields ranging from production optimization to assisted vision. Profile featured in the 3rd issue of GLIMPSE.
Into the abyss of AI
"Before AI and machine learning, we were relatively limited in our ability to explain phenomena with conventional statistics. Deep learning and machine learning make it possible to find correlations or patterns that we weren't able to identify with simple regressions."
Falling into it
"I've always had a great passion for computers, ever since I was 6 years old. A computer was the first thing I asked for before a soccer. My parents were stunned: in the 1980s, computers weren't as common as they are today."
A coder's heart
"I'm in charge of a team of data scientists and data engineers, made up of PhDs in mathematics, engineers or high-level statisticians, but I still code and actively contribute to our projects. It's essential to remain able to participate and maintain an equivalent level of knowledge to be able to be effective on such cutting-edge themes."
The words to say it
"When OpenAI trained its first models, they took everything available on the Internet, a corpus created entirely by humans. But machines have a very statistical approach to content creation. They take the most common words and will end up generating texts from texts they have created themselves. There's a risk that we'll come up against a form of intellectual inbreeding of the model. A bit like in sitcoms, where less than 1,000 words are used for all the episodes."
AI at every stage of the Roquette rocket
"Roquette has always had a strong data culture. The very demanding markets in which we are positioned (nutrition, pharma) have always required us to manage our activities very finely and to ensure perfect traceability thanks to data. The Usine du Futur program was an opportunity to implement the first models using AI to meet our production optimization challenges... And the other functions soon followed: sales prediction and optimization, analysis of vast bodies of documentation, predictive maintenance, talent retention, computer-aided vision... the use cases are numerous!"
KEY DATES
Since 2022
Head of Data & Advanced Analytics, Roquette
2021 - 2022
Head of Data Science, Roquette
Since 2020
Member of the Governance Committee of La Cité de l'IA
2015 - 2021
Future Factory Program Manager, Roquette
2003 - 2015
Projects Director, Rio Tinto
2004
Graduated from SKEMA Business School, Specialized Master in Project Management
