This page is the Apache Arrow developer wiki. If you are involved in building or maintaining the project, this is a good page to have bookmarked. If you are a prospective user of the project, check out user-facing library and API documentation linked to from http://arrow.apache.org/.
The "Arrow columnar format" is an open standard, language-independent binary in-memory format for columnar datasets. It can be used to create data frame libraries, build analytical query engines, and address many other use cases.
Modern machine learning frameworks can leverage technologies we are developing in Apache Arrow, and vice versa.