Scientific computing in Ruby is advancing with libraries like NMatrix, Daru, and Mixed_models. NMatrix provides numerical matrix functionality and initially relied on ATLAS/CBLAS for operations, but a JRuby version using Apache Commons Math is much faster. A proposed general purpose GPU library called ArrayFire-rb would combine Ruby with transparent GPU processing using the ArrayFire C++ library. It could be implemented for MRI using C extensions inspired by NMatrix and for JRuby using Java Native Interface. Benchmarking shows the JRuby version of NMatrix is vastly faster for elementwise operations on n-dimensional matrices compared to MRI. ArrayFire integration could further accelerate applications in fields like bioinformatics, image processing, and computational fluid dynamics