How Do You Structure ML Projects to Reduce Technical Debts?

How do you structure ML projects to reduce technical debts? Where do you start thinking about the problem? Something you can summarize for us? This is an important team organizational principle, because it's very hard if half of the team wants to organize something and the other team is not just doing their own business. The second part is process. As I said, machine learning products are very prone to unspecifiable issues. It means that even if you are very careful, you're going to end up in missing important parts. So you need to have good time box steps on each of them, and then have a way of getting out of this.