This paper describes neural network based approaches to the process of the\nformation and splitting of word-compounding, respectively known as the Sandhi\nand Vichchhed, in Sanskrit language. Sandhi is an important idea essential to\nmorphological analysis of Sanskrit texts. Sandhi leads to word transformations\nat word boundaries. The rules of Sandhi formation are well defined but complex,\nsometimes optional and in some cases, require knowledge about the nature of the\nwords being compounded. Sandhi split or Vichchhed is an even more difficult\ntask given its non uniqueness and context dependence. In this work, we propose\nthe route of formulating the problem as a sequence to sequence prediction task,\nusing modern deep learning techniques. Being the first fully data driven\ntechnique, we demonstrate that our model has an accuracy better than the\nexisting methods on multiple standard datasets, despite not using any\nadditional lexical or morphological resources. The code is being made available\nat https://github.com/IITD-DataScience/Sandhi_Prakarana\n