Built an end-to-end speech recognition for gujarati speech using DeepSpeech2 architecture.

Generation of code-mixed Hindi-English text from Hindi English parallel corpora and addressing theproblems in code-mixed data from linguistic vision.Worked on Moses SMT for translation of code-mixed Hindi-English to either English or Hindi text.

Initially build automatic speech recognition(ASR) system based on traditional architecture for speechrecognition using Kaldi toolkit and tried different acoustic models.Training is done in two different conditions while performing transfer learning. we found the idealnumber of layers to be copied from the base model.

Investigated the optimal placement of controllers on a network by suggesting the location and thenumber of controllers for two network planning scenarios: New SDN network and an existing SDNnetwork that requires change.Calculated the total node weights on each controller. Then we have selected the controller which hasless node weights than mean weight and assign nodes to it.

Developed a Statistical model on WEKA tool for machine learning to classify and Diagnose the stageof Thyroid disease by using various machine learning algorithms.