Although data preparation is critical to building and deploying effective ML-driven systems, data preprocessing steps can introduce subtle data distribution bugs, which may be challenging to detect and resolve. To address this issue, researchers at the University of Amsterdam and NYU have released mlinspect, a library for lineage-based inspection of ML systems that helps data practitioners find and remediate data distribution bugs in real-time. With a library-independent interface, mlinspect can automatically instrument preprocessing operations and run inspections and checks on complex data pipelines (including those that include relational operators, feature encoders, and other operations like normalization or vector concatenation).