Imed MAGROUNE

Research Engineer — AI, Reasoning, Vision, Edge Systems

Compact Reasoning Architectures, Multimodal AI & Edge Intelligence

I design compact AI systems for reasoning, multimodal perception, and embedded intelligence.

Research on small models, structured reasoning, multimodal learning, and efficient deployment on constrained hardware.

Research Projects & Models

Compact models for reasoning, multimodal perception, and efficient deployment

FeynModel

Compact multimodal architecture for vision-language understanding, reasoning, and generation.

View on Hugging Face

LLMEyeCap

Vision-language model for efficient image captioning and visual question answering on constrained systems.

View on Hugging Face

Tiny Reasoning Models

Small-scale reasoning architectures for structured tasks such as puzzles, compositional problem solving, and generalization.

Project Details

Current Research

Selected directions of my recent work

Ongoing

Tiny Reasoning Architectures

Small models for structured reasoning, compositional generalization, and puzzle solving.

Research

Efficient Multimodal AI

Compact vision-language systems for image understanding, reasoning, and efficient inference.

Edge AI

Edge Intelligence & Robotics

Embedded AI systems for perception, decision-making, and deployment on constrained devices.

Publications & Preprints

Papers, reports, and technical releases

In preparation

Tiny Reasoning Models for Structured Generalization

Research on compact architectures for reasoning tasks such as mazes, Sudoku, and structured puzzle solving.

Paper Code
Technical report

Efficient Vision-Language Models on Edge Devices

Experiments on optimized multimodal inference and reasoning under hardware constraints.

Paper Models
Ongoing

Open-source Models and Reproducible Research

Code, model weights, and reproducible implementations for compact AI systems.

Code Models

Selected Projects

Open-source code, workshops, and implementations

Reasoning

Q* — Multi-step Reasoning for LLMs

Illustrated tutorial and implementation combining Q-learning with A*-style search to improve LLM reasoning on the MATH benchmark.

Read Tutorial
Edge AI

FedEdge — Federated Learning on Edge Devices

Privacy-preserving federated learning platform running local LLM inference on Jetson, Raspberry Pi, and standard hardware.

View on GitHub
Workshop

RagLabs — AI Engineering Workshop

Full pipeline workshop: LLMs, vision-language models, RAG, MCP, agent creation, and full-stack AI app development.

View on GitHub
Imed MAGROUNE — AI, Reasoning, Vision & Edge Intelligence

Articles & Tutorials

Technical notes, tutorials, and practical experiments

AIoT
AIoT Tutorial 1: Introduction

May 15, 2023

View Content
AIoT
AIoT Tutorial 2: Sound Processing

June 20, 2023

View Content
AIoT
AIoT: Gerber and GCode Processing

July 10, 2023

View Content

About

I am a Research Engineer working on compact reasoning architectures, multimodal AI, computer vision, and edge intelligence. My recent work focuses on small models capable of structured reasoning, reproducible research, and efficient deployment on constrained systems.

My projects span reasoning systems, vision-language models, embedded AI, and practical research engineering for real-world intelligent systems.

CEA / DRF / IRFU / DEDIP Reasoning Vision Multimodal AI Edge Systems

Platforms & Links

GitHub Hugging Face LinkedIn