CV | Benjamin Kitor

Full Name	Benjamin Kitor
Date of Birth	June 29, 1999

2021-Present

MASc - Computer Engineering

Queen's University, Kingston ON, Canada
2017-2021

BASc - Computer Engineering

Queen's University, Kingston ON, Canada

High-Performance Cluster Computing
- Communication runtimes design, for example, MPI, UCX, and Libfabric
- Collective Communication methods and algorithms
- Software for accelerators like GPUs and DPUs
Deep Learning
- Communication modeling of distributed DL
- Accelerating Model Parallelism/Data Parallelism through efficient communication

Argonne Training Program on Extreme-Scale Computing (ATPESC)
- Was accepted to attend a workshop hosted by Argonne National Lab
- Gained experience running applications at scale on ALCF clusters Theta and Polaris
Process Arrival Pattern Aware Collectives
- Developed a novel method to disseminate process arrival pattern information in distributed memory systems
- Utilizes UCX RMA and atomic operations to minimize synchronization and maximize performance
- Accompanying PAP-Aware allreduce algorithm successfully increased throughput on synthetic Horovod benchmarks
Topology Aware Collective Rank Reordering
- Optimized OpenMPI's Allreduce and Broadcast algorithms through renumbering ranks based on hardware/network topology and collective algorithm
- Measured microbenchmark improvements on systems with up to 128 GPUs and 32'000 CPU core jobs

2020
- Y. H. Temuçin, A. H. Sojoodi, P. Alizadeh, B. Kitor and A. Afsahi, "Accelerating Deep Learning Using Interconnect-Aware UCX Communication for MPI Collectives," in IEEE Micro, vol. 42, no. 2, pp. 68-76, 1 March-April 2022, DOI: 10.1109/MM.2022.3148670.