This talk summarizes some of the work performed during our 3-year Laboratory Directed Research and Development - Directed Research (LDRD-DR) project, titled Machine Learning for Turbulence (MELT). Started in October 2018, the project partially covered 10 staff members, 7 postdocs, several more summer students, and addressed a diverse set of topics related to turbulence and applications in climate and astrophysics. I will survey some of our results on neural network (NN) models of scalar turbulence, embedding hard physical constraints into NNs, Lagrangian dynamics, Mori-Zwanzig structure of turbulence, graphical models and, finally, unsupervised identification of dynamical regimes.