Computer Science PhD Candidate · University of Virginia

Mohammad Sonji

I am a Computer Science PhD candidate in UVA's Insight Lab, advised by Professor Adwait Jog. I study how modern computing systems use GPUs, memory, and interconnects, and how we can make their performance more efficient and predictable.

Previously, I completed an MS in Computer Science at the American University of Beirut with Professor Izzat El Hajj.

Email Google Scholar GitHub LinkedIn CV

Charlottesville, Virginia

Research interests

My work connects architectural behavior with application-level outcomes, with an emphasis on GPU and high-performance systems.

GPU systems
GPU utilization
Memory & interconnects
Unified memory
Application & kernel optimization
Performance modeling
Confidential computing
High-performance computing
Emerging architectures

News

Summer 2026: Passed my PhD qualifying exam.
Jun 2026: Presented Understanding Unified System Memory in Emerging Superchips at YArch 2026.
Jun 2026: Attended ISCA 2026.
Jun 2026: Received travel grants for ISCA 2026 and YArch 2026.
May 2026: Are We There Yet? appeared at ACM/SPEC ICPE 2026.
Apr 2026: HIBISCUS appeared at IEEE ISPASS 2026.

Publications

ISPASS
2026

Hibiscus: End-to-end Architectural Simulation Framework for Hybrid SFQ/CMOS-Memory Compute Systems

Ryan Marsala, Yerzhan Mustafa, Prabhath Tangella, Mohammad Sonji, George Michelogiannakis, Selcuk Kose, Adwait Jog, Mehmet Belviranli

IEEE International Symposium on Performance Analysis of Systems and Software, 2026.

DOI

ICPE
2026

Are We There Yet? Predicting if Executing Applications are Near Completion

Mohammad Sonji, Mohammed Baydoun, Safaa Diab, Amir Nassereldine, Pedro Bruel, Aditya Dhakal, Rolando P. Hong Enriquez, Gourav Rattihalli, Diman Zad Tootaghaj, Gallig Renaud, Barbara Chapman, Fatima Abu Salem, Eitan Frachtenberg, Dejan Milojicic, Izzat El Hajj

ACM/SPEC International Conference on Performance Engineering, 2026.

PDF Slides DOI

ISPASS
2025

Dissecting Performance Overheads of Confidential Computing on GPU-based Systems

Yang Yang, Mohammad Sonji, Adwait Jog

IEEE International Symposium on Performance Analysis of Systems and Software, 2025.

PDF Slides DOI Artifact

iWAPT
2025

Predicting Performance Variability

Mohammed Baydoun, Mohammad Sonji, Pedro Bruel, Dejan Milojicic, Eitan Frachtenberg, Izzat El Hajj

IEEE IPDPSW, International Workshop on Automatic Performance Tuning, 2025.

PDF Slides DOI

GrAPL
2025

Serverless Graph Analytics on Multi-Instance GPU

Mohammad Sonji, Mohammed Baydoun, Aditya Dhakal, Gourav Rattihalli, Dejan Milojicic, Izzat El Hajj

IEEE IPDPSW, Workshop on Graphs, Architectures, Programming, and Learning, 2025.

PDF Slides DOI

Experience

2025 – Present

Teaching Assistant

University of Virginia

CS 6501: GPU Architectures, Spring 2026
CS 4444: Introduction to Parallel Computing, Fall 2025

Jun 2022 – Jul 2024

Research Assistant

American University of Beirut & Hewlett Packard Labs

Worked on performance variability and modeling, heterogeneous computing, and serverless GPU systems.

2021 – 2024

Teaching Assistant

American University of Beirut

Supported courses in computer organization, data structures, programming, and GPU computing.

Mohammad Sonji

Research interests

News

Publications

Hibiscus: End-to-end Architectural Simulation Framework for Hybrid SFQ/CMOS-Memory Compute Systems

Are We There Yet? Predicting if Executing Applications are Near Completion

Dissecting Performance Overheads of Confidential Computing on GPU-based Systems

Predicting Performance Variability

Serverless Graph Analytics on Multi-Instance GPU

Experience

Teaching Assistant

Research Assistant

Teaching Assistant

Community

Service

Honors