Using multi-variate timeseries from multiple groups to train an LSTM model in Keras while using sklearn pipelines

Question

I know the title is a handful. I\'ve done a lot of researching and tinkering around, but haven\'t been able to get the steps quite right to completely achieve the full solution.