Research Experience:

Graduate Research Assistant, UMIACS, University of Maryland, College Park, MD: Advisor: Professor Ramani Duraiswami

I have been working as a Graduate Research Assistant with UMIACS (University of Maryland Institute for Advanced Computer Studies) since January 2007. I have been involved in a number of projects which are listed below:

Speaker recognition: I am currently looking at efficient dimensionality reduction techniques to be used in large dimensional speech vectors. I am developing efficient techniques to perform speaker verification and identification; extending from my work in KRD. Efficient spatio-temporal kriging: I am exploring various techniques for developing an efficient approach for spatio-temporal kriging. I am also developing efficient models to compare kriged trends in atmospheric contaminant concentration with those in meteorological factors, to enable health related predictions from historic data Efficient preconditioners: I am looking at inner-outer Krylov subspace approach for preconditioners, I am trying to solve the preconditioners particularly for radial basis functions. GPU Acceleration of kernel machine learning algorithms I have accelerated kernel algorithms like kernel density estimation, Gaussian process regression, mean shift clustering and ranking using Graphics processor. I have implemented efficient methods for host-device interaction for improving the performance I have developed a generalized extendable approach for speeding up multiple kernels I have extended the GPU based approach to matrix decompositions like LU, QR for kernel matrices. I am currently looking at ways to accelerate linear Improved Fast Gauss Transform on a GPU Fast computation of Kernelized Renyi Distance (KRD): In this research project, I have developed a distance measure based on Renyi entropy for alpha = 2. I have extended the idea to develop a divergence measure and a mutual information measure. The entropic distance computation is further accelerated using FIGTREE and graphical processing units (GPUs). I have the acceleration to develop a KRD-based subset selection approach for Gaussian process regression and object recognition I have extended the KRD measure for evaluating similarity scores to speaker recognition. Dual Estimation using Local Ensemble Kalman Filter I have developed an Expectation Maximization (EM) based dual estimator with Kalman Filter for model and state estimation. The model was estimated using Gaussian Process Regression with data compressed using Informative Vector Machine. I am now exploring the possibility of efficient solutions to non-linear differential equations using a Gaussian Process prior

Summer Intern, National Library of Medicine, National Institute of Health, Bethesda, MD Mentors: Dr. Sameer Antani And Dr. Dina Demner-Fushman

I interned at the National Library of Medicine, National Institute of Health in Summer 2007. I was working on two projects during my internship Developed an algorithm to automatically identify and separate multi-panel images [Report] Improved search retrieval in PubMed by enhancing current image annotation methods Performed shape indexing spinal bones using kd-trees for better database retrieval [Report]

Teaching Experience:

Course: CMSC132 Object Oriented Programming Instructor: Fawzi Emad

I was a Teaching Assistant for CMSC 132: Object Oriented Programming in Spring 2009. My work primarily included Handling lab session for the course Evaluating student projects and home works Grading quizzes and exams