Iman Rahmaty

AI/ML Research Engineer | Distributed Systems & Reinforcement Learning Specialist

Tehran, IR.

About

Highly accomplished AI/ML Research Engineer with a Master's degree from a top-ranked university, specializing in Multi-Agent Deep Reinforcement Learning (DRL) for complex distributed systems and Mobile Edge Computing (MEC). Proven expertise in optimizing resource utilization, enhancing load balancing, and maximizing Quality of Experience (QoE) through innovative algorithmic design and simulation. Eager to leverage advanced DRL approaches and a strong foundation in network optimization to drive cutting-edge research and development in dynamic system environments.

Work

EdgeAI Laboratory

|

Research Engineer

Tehran, Tehran, Iran (Islamic Republic of)

→

Summary

Leads research and development in multi-agent DRL for collaborative computation offloading in partially observable MEC systems, driving scalable and joint decision-making.

Highlights

Developed and implemented advanced multi-agent Deep Reinforcement Learning (DRL) algorithms for collaborative computation offloading in partially observable Mobile Edge Computing (MEC) systems.

Designed scalable and joint decision-making frameworks for heterogeneous agents operating under non-stationary system dynamics, improving adaptability and performance.

Enhanced resource utilization and load balancing by enabling collaborative offloading scenarios, leading to optimized system efficiency and responsiveness.

Pioneered the application of Centralized Training with Decentralized Execution (CTDE) paradigms, fostering cooperative learning strategies among diverse agents.

Performance and Dependability Laboratory (PDL), Department of Computer Engineering, SUT

|

Research Assistant

Tehran, Tehran, Iran (Islamic Republic of)

→

Summary

Conducted DRL-based task offloading research in highly-dynamic and energy-constrained MEC systems, optimizing distributed decision-making for time-sensitive tasks.

Highlights

Developed DRL-based task offloading algorithms for highly dynamic and energy-constrained Mobile Edge Computing (MEC) systems, achieving significant performance gains.

Optimized distributed and independent offloading decisions for time-sensitive tasks, ensuring strict processing deadlines under time-varying network conditions.

Maximized end-user Quality of Experience (QoE) by effectively balancing trade-offs between task completion, latency, and energy consumption based on individual requirements.

Designed solutions that robustly accounted for dynamic workloads at edge nodes, significantly improving system adaptability and efficiency.

Volunteer

Reviewer

|

IEEE Internet of Things Journal

→

Reviewed research articles for the IEEE Internet of Things Journal, contributing to the quality and integrity of published scientific work.

Highlights

Evaluated submissions for technical rigor, methodological soundness, and significant contributions to the fields of IoT and distributed systems.

Provided constructive, detailed feedback to authors, enhancing the clarity, impact, and scientific value of their research.

Reviewer, 27th International Computer Conference

|

Computer Society of Iran (CSICC)

→

Contributed to the peer review process for the 27th International Computer Conference, ensuring high-quality academic content.

Highlights

Assessed conference papers for scientific merit, originality, and relevance to contemporary computer science domains.

Facilitated the selection of impactful research for presentation at a prominent national academic conference.

Head Teaching Assistant, Software Defined Networking

|

Sharif University of Technology (SUT)

→

Supported instruction for 'Software Defined Networking,' guiding students through advanced network architecture principles.

Highlights

Provided in-depth explanations and practical guidance on Software Defined Networking (SDN) concepts and implementations to students.

Collaborated with Professors Ali Movaghar and Dr. Mohammad Hosseini on curriculum enhancement and student assessment strategies.

Head Teaching Assistant, Performance Evaluation of Computer Systems

|

Sharif University of Technology (SUT)

→

Served as Head Teaching Assistant for 'Performance Evaluation of Computer Systems,' supporting student learning and course delivery for over 50 students.

Highlights

Mentored over 50 students in complex system performance evaluation concepts, methodologies, and practical applications.

Assisted Professors Ali Movaghar and Dr. Mahdi Dolati in developing comprehensive course materials and evaluating student assignments.

Conducted supplementary sessions to clarify challenging topics and enhance student understanding of advanced concepts.

Teaching Assistant, Verification of Reactive Systems

|

Sharif University of Technology (SUT)

→

Assisted in teaching 'Verification of Reactive Systems,' focusing on formal methods and system correctness under Professor Ali Movaghar.

Highlights

Facilitated student understanding of formal verification techniques and their application to reactive systems.

Supported Professor Ali Movaghar in delivering course content and evaluating student progress through assignments and exams.

Teaching Assistant, Theory of Machines and Languages

|

Sharif University of Technology (SUT)

→

Provided instructional support for 'Theory of Machines and Languages,' covering foundational computer science topics under Professor Ali Movaghar.

Highlights

Guided students through complex concepts in automata theory, formal languages, and computational complexity.

Assisted Professor Ali Movaghar with lecture preparation, problem-solving sessions, and student engagement activities.

Teaching Assistant, Wireless Networking

|

Sharif University of Technology (SUT)

→

Supported the 'Wireless Networking' course, helping students grasp principles of wireless communication under Professor Ali Mohammad Afshin Hemmatyar.

Highlights

Assisted Professor Ali Mohammad Afshin Hemmatyar in teaching wireless network protocols, architectures, and emerging technologies.

Provided technical support and clarification on challenging topics, improving student comprehension and problem-solving skills.

Education

Sharif University of Technology (SUT)

Tehran, Tehran, Iran (Islamic Republic of)
→
Master of Science Computer Engineering/Computer Networks Grade: 6.08/7

Courses

Thesis: Distributed Computation Offloading based on Deep RL (DRL) in Resource-Constrained MEC

Khajeh Nasir Toosi University of Technology (KNTU)

Tehran, Tehran, Iran (Islamic Republic of)
→
Bachelor of Science Industrial Engineering

Awards

Top 10% M.Sc. Student (Class of 2019)

Sharif University of Technology (SUT)

Recognized as a top-performing student, ranking within the top 10% of the Master of Science program in Computer Engineering.

Nationwide M.Sc. Entrance Exam Rank (55th of 60,000)

National Organization for Educational Testing, Iran

Achieved a top 0.09% ranking (55th out of 60,000 participants) in the highly competitive nationwide Master of Science entrance examination for Computer Engineering.

Nationwide B.Sc. Entrance Exam Rank (Top 1% of 180,000)

National Organization for Educational Testing, Iran

Secured a top 1% ranking (among 180,000 participants) in the nationwide Bachelor of Science entrance examination for Mathematics and Physics.

3rd Position, RoboCup Competition (IranOpen)

RoboCup IranOpen Organizing Committee

Awarded third place in the national RoboCup competition, demonstrating strong robotics and AI application skills.

Publications

QECO: A QOE-Oriented Computation Offloading Algorithm based on Deep Reinforcement Learning for Mobile Edge Computing

IEEE Transactions on Network Science and Engineering

Authored a peer-reviewed paper detailing a novel DRL-based algorithm for optimizing computation offloading in MEC, focusing on maximizing Quality of Experience (QoE).

A Collaborative Computation Offloading Framework based on Multi-Agent Deep Reinforcement Learning in Partially Observable Mobile Edge Computing Environments

Preprint / Under Review

Developed a collaborative computation offloading framework utilizing multi-agent DRL to enhance performance in partially observable MEC environments, currently in progress.

Languages

Persian (Native)

English (Working Proficiency, TOEFL IBT: 108/120)

Certificates

Interactive Learning (DRL Course)

Tehran Institute for Advanced Studies (TeIAS)

Machine Learning and Deep Learning in Python

Udemy

Data Science

Tose'e Higher Education Institute

Advanced Python Topics

Remis Arjang Institute

LPIC-1: Linux Administrator

Anisa Iran Linux House

Skills

Research & Optimization

Stochastic Network Modeling, Network Simulation, Performance Evaluation, Wireless Network Optimization, Distributed Network Optimization, RL-driven Decision-Making, Uncertainty Management, Limited Information Processing, Dynamic System Optimization, Multi-Agent Reinforcement Learning (MARL), Mobile Edge Computing (MEC), Computation Offloading, Resource Utilization, Load Balancing, QoE Maximization, Federated Learning, Collaborative Learning.

Programming Languages

Python, R, Bash, C++, SQL.

Machine Learning & Deep Learning

TensorFlow, PyTorch, Keras, Scikit-learn, Deep Reinforcement Learning (DRL), Dueling Double Deep Q-Network (D3QN), Long Short-Term Memory (LSTM), Actor-Critic Methods.

Data Analysis

Pandas, NumPy, Matplotlib, Jupyter Notebooks.

Frameworks & Tools

Linux, Mininet, Ns-3, Git, LaTeX, Vim, Flask, Visio.

Core Research Domains

Distributed Systems, Wireless Communication, Internet of Things (IoT), Mobile Edge Computing (MEC), Performance Optimization, Multi-Agent Reinforcement Learning (RL), Federated Learning.

References

Prof. Ali Movaghar

Professor of Computer Engineering, Sharif University of Technology (SUT), Tehran, Iran. Email: movaghar@sharif.edu

Prof. Hamed Shah-Mansouri

Assistant Professor of Electrical Engineering, Sharif University of Technology (SUT), Tehran, Iran. Email: hamedsh@sharif.edu

Prof. Mohammad Hosseini

Assistant Professor of Computer Engineering, Shahid Beheshti University, Tehran, Iran. Email: m-hosseini@sbu.ac.ir

Prof. Ali Mohammad Afshin Hemmatyar

Professor of Computer Engineering, Sharif University of Technology (SUT), Tehran, Iran. Email: hemmatyar@sharif.edu

Projects

Multi-Agent Actor-Critic Network Based on CTDE

→

Summary

Developed a hierarchical multi-agent DRL algorithm for collaborative decision-making in non-stationary environments, based on decentralized partially observable Markov decision processes (Dec-POMDPs).

Collaborative Cloud-Edge-End MEC Environment Simulation

→

Summary

Conducted decision-driven simulations of communication-assisted MEC systems, evaluating collaborative computation offloading across cloud servers, edge nodes, and end devices.

Dueling Double Deep Q-Network (D3QN) for Personalized QoE

→

Summary

Developed a D3QN algorithm for distributed decision-making under uncertainty, leveraging both double Q-learning and dueling network architectures. Designed for personalized QoE maximization in dynamic MEC environments based on Markov Decision Processes (MDPs).

Long Short-Term Memory (LSTM) for Dynamic Workload Estimation

→

Summary

Designed an LSTM model to continuously estimate dynamic workloads at edge servers, dealing with limited global information and uncertainty in Q-learning-based decision-making in MEC.

Resource-constrained MEC Environment Simulation

→

Summary

Conducted time-driven simulation of computation and communication processes in queuing-based MEC systems, modeling stochastic system conditions and enabling evaluation of task offloading strategies.

Queueing System Simulation and Performance Evaluation

→

Summary

Performed event-driven simulation and performance evaluation of M/M/1/K queues under FCFS, processor sharing (PS), and discriminatory PS service disciplines.