profile picture

Myungjun (MJ) Son

Ph.D.

Education

Ph.D. in Computer Science

The Pennsylvania State University, USA
2019 - 2025

Master & Bachelor in Computer Science

Kookmin University, South Korea
2010 - 2019

Interests

  • Distributed Systems
  • HPC
  • Cloud Computing

About Theme

Career Profile

Ph.D. in Computer Science with expertise in distributed systems and cloud optimization. Skilled in Python, machine learning frameworks, and high-performance computing (HPC), with a track record of driving cost-efficient, high-performance solutions in cloud-native environments.

Google Scholar

Recent News

  • [08/2025] - Officially graduated with PhD.
  • [06/2025] - Started Research Associate Internship at HPE.
  • [05/2025] - Passed PhD Defense.

Work Experience

Research Associate Intern

Jun 2025 - Sep 2025
Hewlett Packard Enterprise, Milpitas, CA, USA
  • Collaborated with Hewlett Packard Labs’ Systems Architecture Lab to advance distributed systems research and development.

Research Associate Intern

May 2024 - Aug 2024
Hewlett Packard Enterprise, Milpitas, CA, USA
  • Engineered performance and scalability enhancements for distributed HPC systems in collaboration with Hewlett Packard Labs.
  • Designed and implemented a communication framework for distributed HPC systems using ZeroMQ and Mercury in C, improving data handling efficiency across nodes.

R&D Summer Intern

May 2023 - Aug 2023
Takeda Pharmaceutical, Boston, MA, USA
  • Led optimization and automation of data pipelines with the Digital Platforms team, cutting manual processing using AWS analytics tools.
  • Delivered system performance improvements resulting in a 25% reduction in operational costs via data-driven optimizations.

Projects

SpotVerse - Engineered multi-region spot instance provisioning for bioinformatics workflows using AWS CloudFormation, reducing compute costs by 55%. (Python, Bash, CloudFormation)
MicroBlend - Architected an automated resource blending system for microservices, cutting SLO violations by 9% and lowering costs by 11% through telemetry-driven resource allocation. (Python, Microservices, eBPF, Prometheus, Grafana, Jaeger)
Splice - Created a compiler-driven framework for service blending between FaaS and IaaS, reducing workload costs by 32% via optimized workload distribution. (Python, Serverless, Compiler)
ABC2 - Autonomic Big Data Cloud Computing - Analyzed machine learning kernels (Matrix Multiplication) in cloud environments to develop performance model. Reduced matrix multiplication error by 50% using Spark-based distributed computing. (Python, Spark)
DeepSpotCloud - Built a deep learning framework leveraging EC2 GPU spot instances across regions to optimize training costs. Developed a task migration scheduling policy, resulting in a 13% additional cost savings. (Python, Bash, AWS, TensorFlow)

Publications

  • SpotVerse - Enhancing Workflow Efficiency with Multi-Region Spot Instances for Galaxy and Beyond
    • Myungjun Son, Gulsum Gudukbay, Mahmut Kandemir
    25th ACM/IFIP International Middleware Conference (ACM Middleware) (Acceptance Rate - 23%)
  • A Global In-Memory Cache and Computation Tier for DAOS
    • J. Byrne, C. Crasta, ..., Myungjun Son, et al.
    The 9th International Parallel Data Systems Workshop (PDSW’24 in conjunction with SC24)
  • FAAStloop - Optimizing Loop-Based Applications for Serverless Computing
    • Shruti Mohanty, Vivek M. Bhasi, Myungjun Son, Mahmut Kandemir, Chita Das
    15th ACM Symposium on Cloud Computing (ACM SoCC)
  • MicroBlend - An Automated Service-Blending Framework for Microservice-Based Cloud Applications
    • Myungjun Son, Shruti Mohanty, Jashwant Raj Gunasekaran, Mahmut Kandemir
    2023 IEEE 16th International Conference on Cloud Computing (CLOUD) (Acceptance Rate - 21%)
  • Splice - An Automated Framework for Cost- and Performance-Aware Blending of Cloud Services
    • Myungjun Son, Shruti Mohanty, Jashwant Raj Gunasekaran, Aman Jain, Mahmut Kandemir, George Kesidis, Bhuvan Urgaonkar
    The 22nd IEEE/ACM International Symposium on Cluster, Cloud and Internet Computing (CCGRID 2022) (Acceptance Rate - 24%)
  • Modeling Distributed Matrix Multiplication Performance on a Scale-out Cloud Environment for Data Mining Jobs
    • Jeongchul Kim, Myungjun Son, Kyungyong Lee
    IEEE Transactions on Cloud Computing 2019
  • Distributed Matrix Multiplication Performance Estimator for Machine Learning Jobs in Cloud Computing
    • Myungjun Son and Kyungyong Lee
    2018 IEEE 11th International Conference on Cloud Computing (CLOUD)
  • DeepSpotCloud - Leveraging Cross-Region GPU Spot Instances for Deep Learning
    • Kyungyong Lee and Myungjun Son
    2017 IEEE 10th International Conference on Cloud Computing (CLOUD) (Acceptance Rate - 19%)