Service is back to normal.

We've deployed a fix in production. Trainings should now take only a few seconds.

Our data processing infrastructure is running behind which is causing inaccuracies in the training of the bots. No data has been lost, we'll update the incident as soon as the system is back to normal.