Tommaso Grossi

After earning my M.Sc. in mechanical engineering, I embarked on a brief—but incredibly intense—career in the automotive and motorsport industry. I worked as a FEM analyst at Ducati, an autonomous driving specialist at Maserati, and an R&D engineer for the Scuderia Toro Rosso (aka Alpha Tauri / VCARB / any future names I’m not aware of yet) Formula 1 team. It was an exhilarating ride.

But eventually, I realized it wasn’t quite the right path for me. So, I made the decision to return to academia to pursue my passion for research and teaching. While this move may not have been the most financially savvy choice in the short term, it was a refreshing boost for my creativity—something I believe will pay off in the long run.

I completed my PhD in industrial engineering under the guidance of Professor Beghini, an engineering guru with quite a reputation among his students. My thesis showed that many uncertainty quantification techniques used to apply confidence intervals to residual stress measurements are fundamentally flawed and often misuse advanced mathematics. This is a common issue with ill-posed problems. In a nutshell: when dealing with inverse problems, be skeptical of narrow confidence intervals—they’re often misleading. Proving a probability wrong is tricky, especially when the measurements are costly, so the culprits rarely get called out. The full story, along with some potential solutions, can be found in paper and paper.

After completing my PhD, I’ve been exploring how the latest breakthroughs in machine learning algorithms can accelerate structural analysis for mechanical components. The current state-of-the-art approach relies on solving partial differential equations using finite element codes, which treat each analysis as a brand-new problem—without leveraging the insights gained from previous ones, much like a skilled engineer would when working with pen and paper. We can definitely do better than that, and some promising works are coming :)