Yes, you are right on the difference.

Materialize uses incrementally updated materialized view based approach which was influenced by Timely-dataflow paper. Also, there's no disk-based persistence of the materialized views (That is my understanding though).

Pinot on the other hand writes incoming data into columnar files into disk called "segments". Segments are indexed in various way to make the searching faster. Queries are then scattered through the servers who host these segments and results are gathered in real-time.

In summary, Materialize inclined towards in-memory pre-aggregated view model while Pinot favors scatter-gather on indexed data on disk.

Hope this makes sense.