About

Let me introduce myself.

I’m an AI/Robotics engineer with a Ph.D. from UC Irvine (CARL). I build learning-driven, multi-robot systems in ROS2 and increasingly integrate Agentic AI—LLM-powered planners and tools—for language-guided autonomy. Recent work includes a ROS2 platform that trains navigation agents with human data (PyTorch), distributed simulation for multi-agent exploration, and Webots/Gazebo experiments for sim-to-real.
Publications span telepresence navigation and cognitive load (IEEE ICDL 2024) and strategy diversity in robot teams (SAB 2024). Current focus: communication-aware, agentic multi-robot coordination with transformers and tool-use. Outside work I enjoy hiking, camping, gaming, and cat time.

Profile

Full name: Amirhosein Mohaddesi
Role: AI/ML Engineer • Robotics
Website: CARL (Cognitive Anteater Robotics Laboratory)
Email: Amir.mohaddesi@gmail.com smohadde@uci.edu

Selected Skills

For more information download my CV.

95%
Python
90%
C++
85%
ROS2
85%
PyTorch
80%
LangChain / Hugging Face

Contact Me Download CV

More of my experiences.

Education

Direct PhD

2019–2025

University of California, Irvine

Computer Science, Donald Bren School of Information & Computer Sciences
GPA: 3.93/4

Bachelor's Degree

2014–2018

Sharif University of Technology

Hardware Engineering, Computer Engineering
GPA: 18.24/20

Diploma

2010–2014

Shahid Soltani 1 High School, NODET

Mathematics
GPA: 19.8/20

Research Experience

Graduate Research Assistant

CARL Lab

2021–present

Graduate Research – Agentic Multi-Robot Navigation Platform

Built a realistic ROS2 platform to study human-inspired navigation in multi-agent systems; integrated human navigation data for training (PyTorch); added Agentic AI (LLM-driven planning, memory, tool-use) enabling language-guided decision-making and multi-robot coordination; prepared distributed training workflows (DDP-ready); ran Webots/Gazebo simulations with visual/sequence models. Manuscript under review; poster/papers at SAB 2024 and IEEE ICDL 2024.

Navigation and Cognitive Load in Telepresence Robots

We investigated whether autonomous features in telepresence robots reduce cognitive load and improve efficiency. In the autonomous condition, participants navigated more efficiently, performed better in a learning/memory task, and reported lower cognitive load than manual. Presented at IEEE ICDL 2024.

Research Assistant

NMI Lab

2020–2021

DCLL 8-bit quantizer

Implemented an 8-bit spiking neural network for embedded targets and designed an 8-bit quantization technique that reduced power by 12%–18% on MNIST/CIFAR10/20 with only 3%–7% accuracy drop.

Hey, I'm

Amirhosein

Interactive Demo

Career Conversation (Hugging Face Space)