🤖 I translated this article from the website "Les Echos"; it's a very interesting example of how Dassault Systèmes develops AI solutions for its clients, in this case, Valeo.
Valeo employs a battalion of engineers to develop its AI solutions
For over twenty years, artificial intelligence has been an integral part of Valeo’s innovation process and products. The automotive supplier operates two entities that develop AI in parallel—one focused on academic research and the other on industrial solutions.
By Roman Epitropakis – published on September 25, 2025
More than twenty years ago, Valeo was already using artificial intelligence. Its engineers had created a neural network for rear-view cameras to gather the necessary information for pedestrian detection. Today, the automotive supplier applies AI across all its development areas: autonomous driving, of course, but also battery management, light optimization, heat management, and more.
The Valeo Predict4Range is a typical example of such applications. This solution enhances driving range and charging efficiency by utilizing a Virtual Twin of the vehicle, which analyzes real-time data such as outside temperature, traffic conditions, driving style, and even wind speed. The software then optimizes thermal management, increasing range in urban driving by up to 25%, particularly in cold weather conditions.
On long-distance trips, it preheats the battery to optimize charging time upon arrival at the station. Valeo claims it can save up to 1 hour on a 6-hour trip by optimizing both charging time and battery temperature (for example, on long journeys at -7 °C in winter).
Using AI to perform hundreds of tests
To build these solutions, Valeo relies on all the major existing AI models, such as Gemini, Vertex AI, and Mistral. These AIs are trained with Valeo’s proprietary data, all of which is kept strictly in-house, explains Valeo’s R&D Director, Christophe Le Ligné. In the beginning, Valeo had developed its own AI model, “but these were small neural networks, nothing comparable to today’s models like ChatGPT.”
Valeo also collaborates with Dassault Systèmes for all its computer-aided design (CAD) needs, as well as with other companies, such as the Swiss-based Neural Concept, which offers 3D simulations powered by generative AI. The supplier sees a growing range of available solutions and picks the best among them.
AI is also used to break down thought silos in the researchers’ creative process. For example, when designing a cooling circuit, engineers typically start with the most recent model developed and attempt to improve it. “AI doesn’t care about what was done before and proposes completely different circuit layouts,” explains Christophe Le Ligné. “The final circuit developed with AI allowed us to reduce battery weight by 20%, which in turn improved vehicle range.”
Whereas engineering teams might have designed only four or five solutions in a three-year program, AI can test hundreds. It thus optimizes equipment by providing cheaper and more efficient solutions, requiring fewer materials and less weight.
Pursuing academic research and R&D in parallel
“The international pressure on innovation is increasing,” continues Christophe Le Ligné. “The Chinese innovate in twelve to eighteen months, while we in Europe would need thirty-six to forty-eight months.” To stay competitive, Valeo doubled its efforts by launching AI4ALL in 2024. This unit, comprising 200 engineers, collaborates on industrial projects by combining AI experts with specialists from other disciplines to bridge skill sets.
Alongside it is another entity, Valeo.ai, created in 2017. It consists of approximately thirty people who conduct academic research on AI for driver assistance systems. “This team publishes in international scientific journals and participates in conferences on the subject,” notes Christophe Le Ligné. Americans, Indians, Chinese, and engineers from around fifteen other nationalities work together at Valeo.ai.
Roman Epitropakis
