Ali Abbasi

About

I am a PhD student in Computer Science at the University of Southern California (USC), working in the QED Lab under the supervision of Prof. Leana Golubchik. My research interests lie at the intersection of network optimization, traffic engineering in time-varying and satellite networks, machine learning for systems and optimization methods using machine learning. I received my M.Sc. from University of calgary in 2024 and my B.Sc. in Computer Engineering from Sharif University of Technology in 2022.

Research

• Traffic engineering for satellite / WAN networks Predictive optimization and learning-based TE over time-varying topologies, with applications of graph neural networks. Predictive TE formulation combining optimization with GNN-based link availability prediction for satellite and WAN networks.
• Network Flow Interdiction on LEO Satellites Applying optimization and dynamic programming for edge and node removal to interdict facility reachability over tree and bounded-treewidth structures. Dynamic programming and optimization methods for edge and node removal to interdict facility reachability.

Selected Projects

• Federated learning & carbon-aware training Methods such as FedGreen / FedGreenCS and WeiAvgCS for efficient, diverse, and greener FL. FedGreen method adapt model size for carbon-aware, communication-efficient training.
• LLM steering & mixture-of-agents Activation steering for large language models using sparse auto encoders, MoE/MoA-style architectures, and evaluation of their reasoning and coordination capabilities. Experiments with proposer/aggregator architectures and activation steering for reasoning quality.

Publications & Preprints

• • Abbasi, A., Friedman, E., Golubchik, L., Khuller, S., Paolieri, M. (2025). Edge and Node Removal in Tree Networks to Interdict Facility Reachability. (Under Review)
• • Abbasi, A., Dong, F., Wang, X., Leung, H., Zhou, J., Drew, S. (2024). FedGreen: Carbon-aware Federated Learning with Model Size Adaptation. In IEEE ICC 2024 (GreenNet Workshop). (Published)
• • Dong, F., Abbasi, A., Leung, H., Wang, X., Zhou, J., Drew, S. (2024). WeiAvgCS: Federated Learning Model Aggregation Promoting Data Diversity. In IEEE ICC 2024 (ETASN Workshop). (Published)

Teaching

• CSCI 356: Introduction to Computer Systems(USC) – Fall 2024 (Current)
• Previously served as TA for courses such as Data Structures, Computer Networks, and Computer Systems.

Contact

Email: abbasia@usc.edu
GitHub: github.com/aliabbasi7798
LinkedIn: Ali Abbasi on LinkedIn