I developed on a software tool that enables efficient querying on large genetic datasets consisting of many genomes. I researched new strategies and algorithms to increase the efficiency and reduce the memory usage of queries that find substring matches and classify genotypes in these datasets.

I developed embedded software in C++ for Cortex-M microcontroller-based wearable watch devices running on FreeRTOS. Specifically, I productionized a deep learning pipeline using the TensorFlow Lite Micro library to enable real-time inference of hand gestures from electromyography (EMG) data collected by the watch device. I also employed CMake and SCons build systems with Arm GNU toolchains to cross-compile the pipeline code and port it onto the embedded device.

I developed a stochastic mathematical model to predict the timeline of genetic mutation in patients suffering from Huntington’s Disease, leveraging single-cell datasets collected from post-mortem brain samples. Specifically, I used a Continuous-Time Markov Chain to model expansion/contraction mutations in a specific gene across many cells in patients over their lifetime. I implemented the models in R and Java, and leveraged optimization algorithms such as Nelder-Mead and BFGS. I also used compute clusters with qsub to run programs in parallel for a computation speedup. My work contributed to a co-authored paper with the lab, which has been accepted for publication by Cell. Currently, a pre-print is available here: https://www.biorxiv.org/content/10.1101/2024.05.17.592722v1

I participated in the CS/Bio track of the MIT PRIMES research program for high school students. I conducted research under mentor Bob Handsaker from the McCarroll Lab, where I analyzed efficiency and behavior of various clustering algorithms commonly applied to genetic datasets. I later moved onto a mathematical modeling project which I continued as an intern in the lab.