Future Says S3E3: Fostering Data-Driven Innovation and Skillsets
In episode three of this season of Future Says, host Sean Lang spoke to François Deheeger, senior AI fellow at Michelin, to discuss how Michelin is embracing the data and artificial intelligence (AI) revolution.
Like our other guests this season, Deheeger didn’t begin his career in AI. Instead, he earned his undergraduate and master’s degree in mechanical engineering before conducting Ph.D. work centering on uncertainty propagation methods involving machine learning (ML), design of experiments, and surrogate modeling. As his career progressed, he began to see AI, ML, and data spread to other domains and play an increasingly vital role within engineering and manufacturing.
Having spent 12 years at Michelin in various data science roles, he’s now responsible for ensuring that data and AI flourish in the R&D team and beyond. Right now, that means encouraging data-driven mindsets in domains and teams that have historically been on the fringes of data and AI adoption – engineering, manufacturing, supply chain, and others. A good example of this work is in tire production, which Deheeger says has always been a process of working “from the molecule to the vehicle,” i.e., starting from the smallest piece of rubber to sending a completed tire into the world. This process encapsulates starting from material science, moving to structural optimization, and finally creating the connected device.
Since 2015, Michelin has been embedding GPS and accelerometers into cars to understand the real-life duty cycle of its products, but the development of Internet of Things (IoT) and next-generation network technology has brought a whole new dimension to this work. Now, for the first time, companies can connect the development world with the user world and optimize their products thanks to these digital twins. Deheeger is wary to use the term “digital twin” since he knows it can be poorly defined, but he personally saw it in action in two separate use cases – during the virtual build process, and during assets’ in-service life.
The goal of these digital twins in Deheeger’s eyes is to bring the organization and its people into better alignment so teams can communicate and collaborate more closely. They can then make better, quicker decisions to speed development cycles and optimize parts, components, and materials. By doing this, he says, it makes it easier to succeed in both “low-hanging fruit projects and moonshot projects.” These concepts are like what Ford’s Vijayakumar Kempuraj mentioned in episode one, when he discussed both short- and long-term wins. In Deheeger’s view, accomplishing both is key for Michelin. By building toward moonshot projects, it motivates teams and builds a culture centered on innovation. And by simultaneously succeeding in “low-hanging fruit” projects, it produces tangible outcomes that win the support of the team and executives.
An example of each type of project would be AI-driven generative design for new tire sculptures (moonshot) and connecting data from manufactured and designed tires to identify anomalies (low-hanging fruit). As Deheeger says, this is “AI transformation by the book.” If you can generate this momentum, Deheeger says this is when the “Instagrammers” will get behind your ambition – in other words, this is when you can win over people with true industry influence.
Not only is Deheeger a special advisor for the group’s AI transformation, he is still hands-on with AI projects and loves diving into deep learning and probabilistic programming – much like many of his engineering colleagues. “When it comes to AI for engineering, I would say 50% are people like me with backgrounds in mechanical engineering and uncertainty qualification,” Deheeger says. He’s also encouraged that today’s teams aren’t solely composed of software engineers or computer scientists; rather, there’s a combination of mechanical engineers, data scientists, material experts, and others working to expand AI use and deliver better products and processes. As he says, “It’s nice to fall into a group of people trying to make a change.”
To bring in people from assorted domains into AI projects, of course, requires a lot of upskilling. However, for Deheeger, there’s nothing more rewarding – and important – than improving people’s skills and broadening the ways in which they approach and solve problems. Deheeger says this process should be both informative and enjoyable for employees – he tries to incorporate engaging activities like games and videos so people are undertaking something that both captivates them and improves the ways they work. “That’s what we’re trying to do. We want to truly change the way people think and operate day-to-day,” he says. “How do you solve new problems? How do you use data [in new, creative ways]? People have the tools, and we want them to use them in creative ways. That’s the environment I want to build, going beyond single use cases. It’s kind of like a game – and I think it’s fun.”
In the end, his goal is to get organizations working in lockstep so teams can tackle bigger, more complex projects quicker than ever. “[Our goal] isn’t just to be efficient in what we’ve been doing for the past 100 years, it’s to be efficient today and moving forward,” he says. “Tomorrow, with respect to all our capabilities – manufacturing, material science, microsimulation, mechanics, usage understanding, vehicle stuff, etc. – we want to bring that all together so we can tackle different challenges. We’ll be stronger if we can do that – and that’s the challenge.”
