This post doesn’t go too far into the details, but Featuretools has extensive support for time-series datasets. The great part about any manual feature engineering you have already done is that Featuretools can build on top of these features through the use of custom primitives.

If you take a look at this notebook, I use a number of custom time-series primitives building on functions from the tsfresh package. I’d recommend using the features you’ve already made combined with Featuretools to see if you can get a performance gain!