Mohammed Pivezhandi

Mohammed Pivezhandi, Ph.D.

Welcome to my academic website! I recently completed my Ph.D. in Computer Science at Wayne State University (December 2025), working in the Embedded Systems Lab under Dr. Abusayeed Saifullah. My research focuses on data-efficient AI-guided energy- and thermal-aware scheduling for heterogeneous multicore systems, addressing critical challenges in energy efficiency, thermal management, and performance optimization for real-time embedded systems. With a strong background in high-performance computing, reinforcement learning, and computer architecture, I develop innovative solutions for sustainable computing.

About Me

I hold a Ph.D. in Computer Science from Wayne State University (CGPA: 4.0/4.0, 2022-2025), a Master’s in Computer Science/Engineering from Iowa State University (CGPA: 3.76/4.0), and a Master’s in Digital Electronics from Shahid Beheshti University (CGPA: 4.0/4.0), where my thesis on neural networks for human activity recognition was nominated as Best Research in the ECE Department. My career includes research and teaching roles at Wayne State University, Iowa State University, and Shahid Beheshti University, as well as industry experience as a research intern at Moffett Systems, Inc., designing FPGA-based hardware accelerators.

My technical expertise spans Python, C/C++, VHDL, Verilog, TensorFlow, PyTorch, CUDA, and OpenMP, enabling me to bridge software and hardware design. I have mentored over 500 undergraduate students in courses like C programming, computer organization, and machine learning.

Research Overview

My Ph.D. dissertation, Data-Efficient AI-Guided Energy- and Thermal-Aware Scheduling on Heterogeneous Multicore Systems, tackles the growing challenges of energy consumption and thermal management in embedded systems, where power density doubles every three years. My work leverages Hierarchical Multi-Agent Reinforcement Learning (HMARL) and data-efficient RL techniques to optimize energy, reduce temperature hotspots, and enhance performance for parallel DAG workloads on platforms like Jetson TX2 and Intel Core i7.

Research Goals

  • Energy- and Performance-Efficient Scheduling: Minimize energy consumption and makespan for parallel DAG workloads.
  • Thermal-Aware Management: Reduce temperature hotspots to improve system reliability.

Key Contributions

  • HMARL for OpenMP DAG Workloads: Developed a scalable framework achieving 40.95% energy reduction and 49.06% makespan improvement (Published RTSS 2024 WIP).
  • Feature-Aware Task Allocation: Reduced energy by 10% and core temperature by 5°C using statistical learning (Published RTCSA 2025).
  • Zero-Shot Cross-Platform Scheduling: Developed LLM-based feature extraction for cross-platform transfer learning (ArXiv 2025).
  • Few-Shot RL with Flow Matching: Improved data efficiency in reinforcement learning through generative modeling (ArXiv 2025).
  • Uncertainty-Aware GNN Prediction: Integrated evidential deep learning with graph neural networks for robust energy prediction (ArXiv 2025).

My research is supported by National Science Foundation (NSF) funding, and I contribute as a peer reviewer for conferences like AAAI, AISTATS, and ISPA. Explore my Publications and Projects for more details.

Get in Touch

I’m eager to collaborate on cutting-edge projects in AI, real-time systems, and embedded computing.

Visit my CV for a comprehensive overview of my work!