.jpg)
Returning a lease car sounds simple, until someone has to inspect it. Scratches, dents and rim damage all need to be assessed against the Renta standard, a process that is time-consuming, repetitive and dependent on human judgment. That sounded like exactly the kind of problem the AI Lab was built for.
The project combined computer vision, damage detection, explainable AI and industry-specific regulations. Reflections on car paint, inconsistent image quality and heavily imbalanced training data made it the perfect playground for experimentation.
This project started as an eight-week graduation internship in the AI Lab.
Two KdG students worked alongside experienced our Cronos AI specialists to build a working solution for a real operational challenge. Rather than spending their internship on a theoretical assignment, they got their hands dirty on a problem that people inside Cronos face every day.
The biggest breakthrough didn't come from the model. It came from the data. After multiple iterations, the team discovered that improving the dataset had a far bigger impact than tuning the architecture itself. Ergo, data beats architecture.
The result is a working AI-powered inspection platform that:
Most importantly, the system doesn't replace the inspector. It supports the inspector. Human expertise remains in control while the AI does the heavy lifting.
Unlike many student projects, the story didn't end after graduation. Exactly as intended: not as an academic exercise, but as a real solution built around a real-world problem.
.jpg)
%20copy.webp)
