Deep learning is the state of the art in domains such as computer vision and natural language understanding. MXNet is an open-source framework developed through collaboration and contributions of researchers at several universities. It is now the recommended deep learning package at AWS due to its programmability, portability, and efficiency. It is suitable for deployment from multiple GPUs and multiple machines to embedded systems such as smart phones and embedded GPUs. This talk will cover how (i) MXNet has flexible programming paradigms, (ii) MXNet can enable memory-computation tradeoffs and support portability and (iii) MXNet has the highest efficiency on hundreds of GPUs. I will also demonstrate how you can quickly start using MXNet by leveraging preconfigured Deep Learning AMIs (Amazon Machine Images) and CloudFormation Templates on AWS.