Hi, I'm Amir! đź‘‹

Exploring the world of AI & Multi-Agent Systems

From power circuits to intelligent agents, I'm on a journey to make agents communicate more robustly.

My Journey

Amirhossein Karimi

From Circuit Boards to Neural Networks

My academic journey began in the world of Power Electronics at Isfahan University of Technology, where I developed a fascination with circuit design and energy systems. I spent countless hours in labs designing DC-DC converters and optimizing power systems.

But as they say, life is all about evolution. In 2024, I found myself increasingly drawn to the exciting field of Artificial Intelligence. What started as curiosity quickly turned into passion when I discovered Multi-Agent Reinforcement Learning (MARL).

Now at the University of New Haven, I'm researching how to make cooperative agents communicate more robustly, even when faced with adversarial scenarios. It's a thrilling transition that combines my engineering mindset with cutting-edge AI research.

This shift wasn't just about following trends—it was about finding an area where I could make meaningful contributions while constantly learning and growing. The more I work with MARL systems, the more I'm convinced this is where I belong!

2017-2022

BS in Electrical Engineering (Power Electronics) at Isfahan University of Technology

2022-2023

Research Assistant at Industrial Electronics Lab, working on high step-up DC-DC converters

2024-Present

MS in Electrical Engineering at University of New Haven, focusing on robust communication in MARL

Research Interests

Multi-Agent Reinforcement Learning

Exploring how multiple AI agents can learn to coordinate, cooperate, or collaborate and communicate effectively in complex environments.

Robust Communication

Developing techniques to make agent communication resilient against adversarial attacks and misalignment.

Microcontrollers & Embedded Systems

Developing firmware for microcontrollers to create efficient, reliable embedded systems for interactive devices.

Power Electronics

Designing efficient DC-DC converters with reduced switching and conduction losses for renewable energy applications or electric vehicles.

Current Research

Robust Communication in Multi-Agent Reinforcement Learning

At the SAIL Lab (University of New Haven), I work with Dr. Vahid Behzadan on developing mechanisms for robust communication in Multi-Agent Reinforcement Learning (MARL). Our proposed method, Communicative Power Regularization (CPR), enables cooperative agents to maintain effective coordination even under adversarial or misaligned communication.

This work has been accepted for publication at ECAI 2025 and selected for poster presentation at CoCoMARL 2025. I also presented it at Hartford AI Day 2025, where it received insightful feedback from the AI research community.

Featured Projects

MARL Grid-World Simulation

MARL Grid-World Simulation

A customized multi-agent grid-world environment for testing communication protocols between agents under various conditions.

PyTorch Ray RLlib OpenAI Gym NumPy MARL Adversarial Comms scikit-learn TensorboardX GNN
MalConv-based Malware Classifier

MalConv-based Malware Classifier

A real-time malware detection system using deep learning, deployed with AWS SageMaker for cloud-based inference on raw executable files.

PyTorch AWS Streamlit EMBER Dataset MalConv NumPy Scikit-learn LIEF
View Project
Jumping Frog Game

Industrial Power Meter

Designed and implemented the microcontroller firmware for an industrial power meter at Tavan Pajouh Behrad Company, enabling precise measurement and monitoring of electrical parameters in power systems.

C ARM Altium Designer Power Electronics Power Systems
Jumping Frog Game

Frogger Game on ARM Microcontroller

Designed and implemented a complete interactive game on an ARM microcontroller, featuring multiple game levels, game difficulty, real-time graphics, sound effects, and joystick controls in a resource-constrained environment.

C ARM Graphics mikroC PRO for ARM
View Project
High Step-up DC-DC Converter

High Step-up DC-DC Converter

Designed and prototyped a novel converter with reduced switching losses and improved efficiency for renewable energy applications.

Power Electronics Altium Designer PSIM MATLAB
View Paper

Publications

2025

Robust Coordination under Misaligned Communication via Power Regularization

Nancirose Piazza, Amirhossein Karimi, Behnia Soleymani, Vahid Behzadan, Stefan Sarkadi

European Conference on Artificial Intelligence (ECAI) 2025

2024

High step-up DC-DC converter with low conduction losses and reduced switching losses

Zahra Akhlaghi, Amirhossein Karimi, Ehsan Adib

IET Renewable Power Generation, January 2024

Skills & Tools

Programming

Python
PyTorch
TensorFlow
NumPy
Pandas
Scikit-learn
Ray RLlib
JAX
C
C++
MATLAB
HTML
CSS
JavaScript

Tools & Platforms

AWS Git WordPress Streamlit OpenAI Gym PettingZoo Hugging Face Stable Baselines3 LaTeX

EE & Hardware

Altium Designer PSpice OrCAD Proteus AutoCAD ARM MicroC PRO PSIM MATLAB Simulink Verilog CodeVisionAVR ATMEL Studio DIgSILENT STM32CubMX Keil ISE Isim Xilinx PlanAhead Modelsim PSCAD PVSOL DIALux

Get In Touch

Let's Connect!

I'm always open to discussing research, collaboration opportunities, or just chatting about AI and reinforcement learning.