From missionary in Beirut, Lebanon to scientist working with the most powerful supercomputer in the world, I have had the honor of working on brilliant projects with incredible people. My family, mentors, and faith play significant roles in my success and perserverence. Thank you to all who believe and encourage me.
Currently, I am a graduate student at Northeastern University studying Computer Engineering. To fund my degree, I received a research assistantship at Goodwill Computing Lab and a GEM fellowship from Oak Ridge National Laboratory.
My research at Goodwill Computing focuses on analyzing and improving file system I/O on large-scale high-performance computing (HPC) systems. As I/O demand of scientific applications increases, identifying, predicting, and analyzing I/O behaviors is critical to ensure parallel storage systems are efficiently utilized. To tackle this particular performance barrier, I developed a unique methodology that can be used to perform temporal and feature analyses that detects interesting I/O patterns in production HPC systems. My work also delves into the implications for operating and managing new cutting-edge supercomputer systems. My first paper as first author is based on this research and set to be published at SC '21.
In addition to my university research, I am involved with the CANcer Distributed Learning Environment (CANDLE) project at Oak Ridge National Laboratory. As part of the Exascale Computing Project (ECP) and in collaboration with the National Cancer Institute (NCI), the CANDLE project is developing methodology to train deep learning models that accurately classifies and analyzes pathology reports on HPC systems. The CANDLE project has historically used computation resources at the Oak Ridge Leadership Computing Facility (OLCF) such as Summit and Titan. Soon, OLCF is deploying Frontier, the first computing system to exceed exa-FLOPs. To enable users to prepare for using the new exascale resource, OLCF deployed Spock, a smaller system with similar architecture to Frontier. My current role in the CANDLE project is to help transition to Frontier, which uses AMD MI100 GPUs. This is a major shift in architecture from the previously used Summit supercomputer which uses NVIDIA Volta V100 GPUs.
My research at CANDLE accesses the potential performance of Frontier's architecture. I use Spock to analyze the expected performance of BERT, a transformer-based machine learning technique for natural language processing pre-training, on Frontier. This performance is compared with Summit and Perlmutter, a new supercomputer at the National Energy Research Scientific Computing Center (NERSC). This research is ongoing and I expect a publication in early 2022.
In my freetime, I enjoy traveling, sailing, cooking, hiking, reading, grabbing a nice cocktail, and writing poetry.