This article is about reusing existing C/C++ CUDA implementation in Python. The rationale behind doing this is, doing fast prototyping in Python while CUDA does most of the heavy lifting in C/C++. We are going to use shared objects to do so.