@wmikolajczyk and I are working on Kaggle Predict Future Sales competition.
We try to beat approaches that we've seen in this competition so far. In particual, we want to try how CatBoost compares to more popular XGBoost, and how entity embeddings can be used for categorical features in this dataset.
So far, our preprocessing is largely borrowed from @szhou42 kernel, but we don't use all the techniques. It can be found here: https://www.kaggle.com/szhou42/predict-future-sales-top-11-solution