I am currently a Principal Data Scientist at Fidelity Investments, working on speech and text data to design a conversational AI technology. Previously, I was a Speech Research Scientist in 3M/M*Modal, where I was part of the research and development group for speech enhancement and recognition for medical applications. Before that I was a Postdoctoral Associate in the Department of Electrical and Computer Engineering (ECE) at Duke University, where I was working with Prof. Vahid Tarokh. I obtained my Ph.D. degree in Electrical and Computer Engineering (ECE) from Iowa State University (ISU) under the supervision of Dr. Chinmay Hegde.

My research interests lie in the intersection of machine/deep learning, statistics, signal processing, and computational mathematics. In my postdoc, I was working on different projects, including deep neural network compression, neural architecture search, few-shot learning, time series analysis, inverse problems in designing meta-materials, radar signal processing, speech recognition, and audio processing. As reflected in my CV, I have established an interdisciplinary research approach in fields ranging from optimization theory to machine/deep learning, and statistics. I am also interested in applying machine/deep learning algorithms in speech recognition and understanding, natural language processing, computer vision, and other disciplines.

In my Ph.D. thesis, I have accomplished various projects, including sparse signal approximation and compressive sensing, analyzing inverse problems such as image denoising and nonlinear demixing of the structured signals, learning latent variables in probabilistic graphical models, and theoretical understanding of neural networks. In all these projects, the main focus was on developing fast and computationally efficient algorithms. As a result of my research, I could publish several papers in various journals, conferences, and workshops. Also, during my internship at Technicolor AI Research Lab, I have worked on different projects in deep learning, including Generative Adversarial Networks (GANs) for solving inverse problems in image processing, designing efficient defense mechanisms against adversarial attacks in adversarial machine learning, and speeding up deep learning computations in the pre-trained models for real-time applications such as object detection in videos.

In addition to doing research, I have a great hands-on experiences in training, debuggig, testing, visualizing, and deploying machine learning models, and I have worked with many different technology, including but not limited to Python, Numpy, Pandas, Streamlit, PyTorch, PyTroch Lightning, Hugging Face, AWS, Sagemaker, Athena, Docker, Bash/Linux, SQL, Git, and Latex. Moreover, I have a valuable experience in mentoring both undergrad and grad students for their class projects and Master/Ph.D. thesis. I have also taught a variety of courses. In particular, I am teaching a machine learning course as a AI mentor program in Fidelity. During my postdoc at Duke, I was also an instructor for the Multivariable Calculus course in the Department of Mathematics and the head TA for courses such as Signal and Systems, and Advanced/Introductory course in Deep Learning in the ECE Department. Moreover, in my graduate studies, I have served as a teaching assistant for courses such as Signal and Systems, Deep Learning, and Electrical/Electronics Circuits.