Increasingly, data practitioners are training ML models on multi-modal datasets and iterating on data and features to improve model performance. In this context, researchers from Stanford University have released Meerkat. Meerkat is centered upon a simple columnar data abstraction, DataPanel, which can represent columns of arbitrary types (e.g. images, graphs, time-series), and which loads high-dimensional data lazily (thereby supporting larger-than-RAM datasets). It is designed to make data inspection, evaluation, and training of multi-modal and other models more efficient and robust. For example, DataPanel makes it easy to store model predictions and possibilities by supporting map, update, and filter operations.