SystemML enables declarative machine learning on Big Data in a MapReduce environment. Machine learning algorithms are expressed in DML (Declarative Machine learning Language), and compiled and executed in a MapReduce environment. DML is a declarative high-level language, that significantly improves the productivity for implementing machine learning algorithms. DML exposes several constructs including linear algebra primitives that constitute key building blocks for a broad class of supervised and unsupervised machine learning algorithms. It also incorporates machine learning constructs such as cross validation. SystemML implements optimization techniques to generate low-level execution plans for MapReduce. The optimizations are based on data and system characteristics. As such, SystemML provides orders of magnitude performance improvements for varying data sets compared to algorithms directly implemented in MapReduce. Our scalable operators are implemented in Hadoop, an open-source MapReduce implementation.