They are very different, Julia is a multi-paradigm dynamically typed language at it's core, with heavy inspiration in the Lisp family. The type system is designed for the purpose of specialization and dispatch, meaning if you add two variables, the compiler will look at hundreds of possible implementations of addition for the one that best fit the combination of arguments, with increasingly more generic and less performant options as fallback until it's unable to match any implementation and fails (after running the program).

Dex is a functional statically typed language inspired by the ML family. That means type errors will be caught before even compiling the code, which gives a guarantee that once it starts running there will be no case of failure because there is no implementation of a method for given types.

The similarity between the two languages seem to be mostly the high performance powerful native multidimensional array support.

Dex looks very different from Julia, particularly in the type system. Julia is fundamentally similar to a cleaned-up Matlab or R, with some extra tricks on top (e.g. multiple dispatch and macros), and a very good and polished implementation.