"Pandas alternative" kind of undersells it -- it's <i>drastically</i> faster and supposedly has a much more intuitive interface. The limits of how long you can keep doing things entirely in-memory (and postpone the move to something like Spark) get higher and higher.