We present several applications of machine learning techniques in Finance and show some details on a calibration project and a risk management project (in presence of frictions). Several theoretical insights why machine learning can make a difference in calibration, risk management or filtering are presented.