Machine learning at Pirelli R&D
A summer in the R&D team at Pirelli's motorsport tyre manufacturing facility in İzmit, working on a machine learning problem drawn from the production environment.
What I can say
I spent the summer of 2026 with the R&D team at Pirelli’s motorsport tyre manufacturing facility in İzmit, working on a machine learning project defined by, and deployed into, the production environment.
Motorsport tyre manufacturing is an unusually demanding setting for applied machine learning. Volumes are low, tolerances are tight, every unit matters, and the physical processes involved are not the kind that come with clean labelled datasets attached. Working there meant treating the factory as the source of truth and building around what it could actually provide, rather than around what would have been convenient.
The work ran end to end: understanding the problem from the people who own the process, getting the data into a usable state, building and iterating on the model, and getting the result to run reliably outside a notebook, on real hardware, in a real facility, in the hands of people whose job is not machine learning.
What I cannot say
Everything else.
The problem definition, the data, the model architecture, the performance figures and the deployment details are all confidential, and will stay that way. I have deliberately kept this page vague rather than write something that sounds impressive at the cost of the confidentiality I agreed to.