Predictive modeling

I have a dataset with 20k observations and 147 features and goal is to use these features to predict a binary outcome. There will also be a test data of the same format of 29k obs for you to predict the outcomes. The datasets are well organized in csv formats. The challenge is: 1. some features are numeric and some are categorical; 2. many features have missing values. Need some data imputation for both train and test datasets; 3. some kind of model selection is necessary given the large number of features, for example Lasso. The project is due in 48 hours. On day 1, I would like to see a functional model, and the next day can be devoted to improving the model. For a very experience data scientist, the whole project won't take more than 8 hours. Python 3 is the preferred programming language. Obviously, you will need to use Numpy, Pandas, Scikit Learn. Ability to use Tensorflow and/or Keras is a strong plus.