Join one of the architects of CUDA for a step-by-step walkthrough of exactly how to approach writing a GPU program in CUDA: how to begin, what to think about, what to avoid, and what to watch out for. Building on the background laid down in the speaker's previous GTC talks "How GPU Computing Works" and "How CUDA Programming Works" (although there is no requirement to have seen them), we'll start from first principles to apply everything we know about parallel and GPU programming to create a CUDA application from scratch.