Brown University

Postion: Research Intern
Interned in Laboratory of Neural Computation and Cognition, under supervision of Professor Dr. Michael J. Frank.
Wroked on implementing reinforcement learning in spiking neural networks and studying noise and signal correlations in teh network. Aim of the internship was to combine two literatures: RL in spiking networks and Neural Correaltions. The research work still in progress.
Tools used: Python, Matlab, Google Colaboratory, Brian
Concepts: Spiking Neural Networks, Neural Correlations, Reinforcement Learning

Defense Research and Development Organisation (DRDO)

Postion: Summer Trainee
Developed a classifier system using convolution neural networks for detecting and identifying underwater objects using hydrophone data recordings by a submarine. Also, was instrumental in the development of triplet hydrophone sensor, which can calculate the direction of arrival while recording the underwater signals. Supervised by Scientist Sanjeev Kumar.
Tools used: Python, Keras, Matlab
Concepts: Convolution Neural Networks, Signal Processing, COmmunication Systems, Digital Circuits and Systems

eXtended Learning Systems (XCS)

Postion: Research Assistant
Created a rule based Artificial Intelligent (AI) agent to learn the optimal game playing strategy for Othello (a combinatorial board game, very popular in AI research) by evolving the rule population through a combination of Reinforcement Learning and Genetic Algorithm. The agent showed a high winning percentage when tested against other state of the art Othello play-ing AI agents. Supervised by Dr. Swati Aggarwal, Professor of Artificial Intelligence in NSIT The project resulted in a paper which has been accepted for presentation at the IEEE CEC 2018(Brazil) and for publication in the conference proceedings published by IEEE
Tools used: Java, Python, Latex
Concepts: Reinforcement Learning, Genetic Algorithm, XCS

Computational Linguistics

Topic:Understanding Language Dependency on Emotional Speech using Siamese Network
Aim to study the susceptibility of emotions, expressed in speech, to different languages. The Project introduces a novel Deep Convolution Siamese Network (DCSN) for determining the similarity between speech samples.The results provided a conclusive proof for the model to be used for the study of different psychological aspects of emotional speech. Paper is under submission in EMNLP-2018 to be held in Belgium.
Tools:Python, Keras
Concepts: Deep Learning, Siamese Networks

Emotion Morphing in Speech

Introduced a deep learning approach to transform emotions in natural speech without altering acoustic features. Developed a deep Convolutional Neural Network (CNN) based Encoder and multiple shallow CNN Decoders for morphing emotions. Paper is under submission in Interspeech, 2018.
Tools:Python, Keras
Concepts: Deep Learning, Signals and Systems