Mohsen Raoufi | Portfolio

About Me

Hey there!

I am an AI Engineer and Robotics Specialist interested in learning, control, and dynamics of complex systems. I am driven by a simple belief: to truly understand a system, you have to build it yourself. This principle has shaped my decade-long practice of designing, implementing, and deploying intelligent systems, from high-performance controllers and decentralized learning algorithms to embodied robotic collectives. With a background in dynamics, state estimation, and control, I specialize in the full lifecycle of intelligent systems: from mathematical modeling and data-driven simulation to real-world deployment.

These days, I am extending my expertise into AI-driven automation and full-stack development. My focus is on managing the entire development cycle—integrating data science backend logic with functional front-end interfaces and testing frameworks. I have become the person people go to for solving intricate challenges end to end.

Skills

Engineering & Dynamical Systems

Matlab & Simulink

State Estimation

Sensor Fusion

Model-Based Design

Robotic Systems Integration

Gazebo

Webots

Embedded Systems

Systems Programming & Interactive Tools

Python

C++

Julia

CUDA

Qt

Interactive Dashboards

D3.js & Observable

JavaScript

Data, Modeling & Scientific Computing

Data Analysis & EDA

Statistical Modeling

SQL & BigQuery

Agent-Based Modeling

AI, Learning & Optimization

Bayesian Methods & Probabilistic Models

Reinforcement Learning

Artificial Neural Networks (ANN)

Graph Neural Networks (GNN)

Optimization Algorithms

Agentic AI (Learning ...)

LangChain / LLM Orchestration

Retrieval-Augmented Generation (RAG)

Model Context Protocol (MCP)

n8n (Workflow & Automation)

DevOps, Infrastructure & Tooling

Docker

CI/CD & GitHub Actions

Git & Version Control

HIL/SIL

System-Level Perspectives & Concepts

Collective Intelligence

Multi-Agent Systems

Control Theory

Network Science

Complexity Science

Selected Projects

<LARS>: Light Augmented Reality System

Augmented Reality
Human-Robot Interaction
Software Development
Robotics
Collective Systems
C++
Qt
CUDA
Computer Vision
End-to-End System Design
OpenCV
Data Analysis

<LARS> is a standalone framework engineered to provide a seamless interaction between physical collectives and virtual environments. As an end-to-end pipeline, it integrates high-speed detection, real-time tracking, and dynamic projection into a single, cohesive architecture. Built on a robust Model-View-Controller (MVC) pattern, the system enables researchers to bypass complex hardware setups and jump straight into experimentation. The core of <LARS> lies in its marker-free tracking engine.

Heterogeneous Collective Opinion Dynamics

Collective Systems
Network Science
Modelling & Sim
AI & ML
Bayesian
Probabilistic Modeling
Multi-Agent Systems
Python
Data Analysis
Complex Systems

This collaborative project introduces a Bayesian framework for opinion dynamics, moving beyond deterministic models to formally model uncertainty within social networks. By treating opinions as probability distributions rather than fixed points, we explore how agents update their beliefs when faced with noisy or conflicting information. The research investigates how heterogeneity in network topology (centrality) and quality of information (certainty) changes the dynamics collective consensus. In this work we combined network science with probabilistic inference to provide a mathematical foundation for understanding how opinion dynamics evolves in complex, decentralized systems, offering key insights for both social modeling and the design of resilient multi-agent AI.

Emergence and Breaking of Echo Chambers in Social Systems

Collective Systems
Modelling & Sim
Python
AI & ML
Agent-Based Modeling
Probabilistic Modeling
Network Science
Multi-Agent Systems
Complex Systems
Data Analysis

This research studies the challenge of echo chambers in networked systems, where homophily—the tendency to interact only with like-minded peers—prevents consensus and drives polarization. Using agent-based simulations and complex systems analysis, I modeled how these clusters emerge and explored a novel strategy to break them: 'Messengers', by utilizing a Dichotomous Markov Process. This work provides a scalable modeling and simulation framework for understanding social dynamics and multi-agent systems.