How would you design a pipeline that supports both batch reporting and near-real-time analytics at TaskRabbit?
