Programming for scientific applications requires to combine two contradictory requirements: rapid prototyping and fast code execution. This is classically solved by the two-language approach where one language is used for prototyping and the second for speed. We have historically tested Matlab/Mex and Python/Cython approach. However, writing in two languages represents a major maintenance issue. Julia language was designed as a solution to this problem at MIT in 2009. Since its public release in 2012  it has grown into a major platform. The major key to success is the new programming paradigm based on multiple-dispatch that  allows far more code reusability than procedural or object-oriented approaches.  In this talk, I will show elementary properties of julia and demonstrate its advantages in comparison with Matlab and Python