An overview of model development at NASA GISS is focused on parameterization schemes relevant to cloud processes. Wholesale replacement of boundary layer turbulence and stratiform cloud schemes was accomplished with an emphasis on comparing ModelE performance in single-column model mode with large-eddy simulation results over a range of conditions: neutral and stable clear-sky boundary layers; warm cumulus, stratiform, and transitional maritime clouds; Arctic cold-air outbreak and decoupled mixed-phase clouds; synoptic cirrus; and continental and tropical deep convection cases. Forty cloud-related tuning parameters were then simultaneously varied over estimated uncertainty ranges using a machine-learning approach constrained by roughly twenty satellite data sets. A small ensemble of equally likely physics obtained from the machine learning process and one “modeler's choice” tuning with substantial changes to the convection scheme are undergoing more detailed evaluation against satellite and ground-based data sets using improved and new simulator codes. Additional large-eddy simulation cases are under development using new field campaign data sets with an emphasis on better constraining aerosol-cloud interactions in warm and ice-containing clouds.