I read and understood several parts of the documention of the Cuda-Convnet project you posted, such as Data and LayerParams, and the rest I just quickly skimmed through and will read them thoroughly once I need it. The documentation seems quite easy to understand. But first I will need to know what add-ons exactly you need to get developed. I have experience with convolutions as I have used convolution formulae to match image parts in a few projects. I also have good experience in Cuda parallel programming. I have been doing Cuda Programming using Visual Studio 2012 C++. However, I saw your project requires python, which is not a problem, I can download and install appropriate packages for CUDA programming in Python. I won more than 35 coding competitions, and I usually write fast and bug free code. Pay me only when you see my code working. Let me know if you wanna give it a shot with me. Also, since I dont know what exactly you need to implement, I cant tell you the price as of now. I will let you know the price once u send me the details.