I am currently a Master of Science in Artificial Intelligence (MSAI) and Engineering Management (Minor) student at Northwestern University. I have diverse experience in AI projects, machine learning tools, and deep learning frameworks, having developed novel AI techniques in my previous role as an AI Engineer (R&D) at Zoho CRM (SaaS) and have previously worked as a Computer Vision (AI) Research Intern at the Amrita CREATE + UMANG Indian e-Governance Services (collaboration) and have also worked as an ML Intern at a tech startup called Easycrop.
I also published research in machine learning and IoT during my undergraduate program and served as a Teaching Assistant for the MBAI program at Kellogg. My academic and practical involvement in AI projects, along with my commitment to the academic and practical application of AI knowledge, position me well to contribute effectively.
Imagine you lost your spectacle and the world around you is completely blurred out. As you stumble around, you see a small animal walk towards you. Can you figure out what it is? Probably yes right? In this situation, or in foggy/night-time conditions, visual input is of poor quality; images are blurred and have low contrast and yet our brains manage to recognize it. Is it possible to model the process? Does previous experience help?
Maintaining the Brain Functional Connectivity (BFC) in the gambling task using the HCP fMRI data-set.
Are there any regions and networks in the brain which are more affected by the Win and Loss Event conditions corresponding to someone’s decision? To analyze data-set being preprocessed by HCP© for NMA-2021 that they recorded using fMRI is to analyze in this project to maintain connectivity and correlation of the areas which are involved in the task.
Deep fakes are a machine learning approach in which a person’s resemblance is changed in an existing image or video. Deep fakes have become a social problem since they allow anyone’s image to be co-opted, putting our ability to believe what we see into question. We create a GAN to generate deepfakes in this project.
Time Series Forecasting based on Ethereum Cryptocurrency Stock prices data with different Machine Learning and Deep Learning models
A machine learning-based time series analysis was used to anticipate the market price and stability of Ethereum in the Crypto-market. Time-series analysis can forecast future price fluctuations in Ethereum. For time series analysis, we employed LSTM, moving averages, ARIMA, and FBProphet as machine learning approaches.
In this project, we use the PyTorch library to create the Reinforcement Learning method using the Deep Double Q-Network (DDQN) algorithms. We demonstrate how the recently developed Double Q learning (DQN) technique, which combines Q-learning with a deep neural network, may be utilised to create an agent that assists in completing levels in Super Mario Bros.
In this project we present the use of Splinter and SpanBERT models to showcase and try to solve the question-answering problem both in the closed domain and open domain as well, where the dataset used was (open domain) is SQuAD 2.0 and also a separate dataset (COVID dataset) has been generated by us for the splinter model (closed domain).
Speech Word-Recognition with Hidden Markov Model (HMM)
The speech recognition system implemented during this project trains one hidden Markov model for each word that it should be able to recognize. The models are trained with labeled training data, and the classification is performed by passing the features to each model and then selecting the best match.
With the persisting scenario of increased privacy and security breaches in chatting and social media platforms, there’s been a sense of insecurity stirring amongst people globally. Decentralized applications use peer-to-peer networks, this eliminates the possibility of network failure due to central node failure. Blockchain functions as an immutable ledger that allows for decentralized messaging.
With the rapid development in deep learning, more powerful tools, which are able to learn semantic, high-level, deeper features, are introduced to address the problems existing in traditional architectures. This YOLO-V4 behave differently in network architecture, training strategy and optimization function, etc.
In this project the basic aim was to try and fit in the overlay of a subject video onto another video containing the respective frame for playing the latter subject video. The solution to this problem was achieved by making the user define the four corners of the frame and then the previously mentioned corners using the Lucas-Kanade Optical Flow function in OpenCV were being kept track of.
Data Structure and Algorithm (DSA)-based project which performs basic Phone Directory function abilities using the concepts of Binary Search Tree and Object-Oriented Programming using JAVA.
Created a Movie Recommender System that seeks to predict or filter preferences according to the user's choices. Used Spark MLlib library for Machine Learning provides a Collaborative Filtering implementation by using Alternating Least Squares.
Worked on an existing research project called deep image prior which is based on image restoration which solves all image inverse applications with a new deep-prior algorithm.