Natural Machine translation has shown great strides in the past few years thanks to introduction of encoder-decoder architecture and attention mechanism. A great deal of research by both industry and academia has enriched the field. Sockeye is an open source framework for Sequence-2-Sequence models with attention that is developed by Amazon. Sockeye is implemented using Apache MXNet, a fast DL framework, optimised for distributed training and efficient inference. This presentation provides a brief history of Language Models and translations as provides an example of NMT using Sockeye.