This document describes a parallel, distributed-memory framework for comparative motif discovery. The framework uses a MapReduce implementation to exhaustively explore sequence words in a gene family dataset from multiple organisms. Keywords are enumerated using a generalized suffix tree and scored for conservation across organisms. Significant motifs are identified based on branch length scores and statistical evaluation in comparison to random permutations. The framework scales well using cloud computing resources. Further work is needed on post-processing and filtering of motif output.