#bbuzz 17: Distributed and Native Hybrid optimizations for Machine Learning Workloads