Hi, I'm Arjun. I'm an electrical engineer and a nuclear engineering Ph.D. student at NC State. Under Dr. Mihai Diaconeasa, my research broadly falls under the Probabilistic Risk Assessment (PRA) umbrella, with applications in the aerospace and nuclear industries.
I aspire to keep my work open and accessible. Sources and documentation for my current and past projects are freely available. Please email me if you find any of my published work behind a paywall.
I completed my Masters in August 2023. As part of my thesis, during the summer of 2021, I interned at Idaho National Laboratory (INL), working on the reliability of Autonomous Control Systems for fission batteries.
Until August 2020, I worked as a research engineer at The B. John Garrick Institute for the Risk Sciences, UCLA on a collaborative project with NASA's Jet Propulsion Laboratory (JPL), led by Dr. Ali Mosleh. I served as the PRA software development team lead for part of that time.
In 2017, I completed my undergraduate research, a collaboration with UCLA SmartHealth, focused on building a cloud-deployed, integrated machine learning framework for remotely monitoring geriatric health outcomes using smartwatches.
From 2014 through 2015, I worked on embedded systems research in Dr. Majid Sarrafzadeh's lab in the Computer Science department at UCLA.
C++ implementation of a 2D neutron diffusion equation solver and supporting BLAS library. Implemented using the discrete ordinates and finite difference methods.
Convex polygon detection using linear hough transform. Polygon detection framework for bitmap images written in C that implements Otsu's thresholding and a Deriche edge detector for blurring, calculation of magnitude and gradient direction, non-maximum suppression and hysteresis thresholding. As a final step, it performs a Hough transform to detect the number and orientation of each polygon edge in the source image.
Indoor localization on an Android device using WiFi 2.4/5GHz, Bluetooth and GSM/UMTS/LTE through supervised learning. Uses a kernelized Naive-Bayes classifier implemented using the Weka machine-learning framework.
An OpenCV implementation of an Analog Utility Meter Reader. Image processing code written in Python, server-side visualization implemented using Node JS and MongoDB.
Vowel recognition engine, written in C using PortAudio (compatible with OS X, Windows, Linux). Recognition performed by performing k-means clustering on MFCC features in time-segmented integer audio samples.
Real-time detection using supervised learning on Android Wear. Distinguishes between walking, running, lying down, sitting, standing or inactive. 256 extracted features include energy & entropy in time & frequency domains. Classification using deep neural networks performs at ≥85% accuracy in real-world scenarios.
Developed for NASA JPL, performs Kaplan's two-stage Bayesian Updating procedure, for arbitrary probabilistic distributions.
Developed for NASA JPL, computes time-to-failure estimates for semiconductor devices using physics-of-failure modeling techniques.
Developed for Pacific Gas and Electric (PG&E), California, real-time health-monitoring and prognostics, corrosion modeling, and optimal mitigation actions for gas and oil pipelines.
Developed for Pacific Gas and Electric (PG&E), California, real-time monitoring of wildfires due to power-grids, including fire-spread and egress.
Developed for the Japanese Nuclear Regulatory Authority, a Human Reliability Assessment (HRA) tool built on a cognitive human response model.