Public Repos

I work in a large company, various departments have different public Repos. Each of us started using a famous ML framework and this has resulted in each area publishing similar samples (MNIST, CIFAR-10, Census, Iris, etc.) which has led to a lot of code duplication, different solutions for same problem, lack of standardization and maintenance. (Example: Each department needs to update their samples to a more recent version, others do not maintain or have already deprecated similar samples). Starting a new initiative is above my manager and my pay grade. I started contacting some others teams and they seem interested in collaborate by using our samples. What would be a good starting point and strategy long term? Anybody face similar problems?

​