Developing an effective architecture for speaker identification system which performs better on different length of utterances. Better representation of the utterance to uniquely identify the speaker, Neglects the non-speaker related features, and Leverage past and future information of utterance to improve the performance are the Main motives of this project.