Hello! Is it possible to train a model to predict daily time series e.g. 90 days in the future and evaluate the monthly aggregated predictions in addition to the daily ones using Nixtla losses?

During training, you can only compute the loss on one scale. In your case, it seems that you have daily data, so the loss will be for the daily predictions. However, once you have the predictions, you can aggregate them and compute metrics on both daily and monthly scales. I recommend using

evaluation

from utilsforecast for that.