Houssem Eddine

AI Engineer · Medical Imaging · Agentic Systems · Computer Vision

📍 Ariana, Tunisia

📞 +216 46 210 801

Open to relocation · Remote-friendly · Available immediately

Clinical-grade Medical AI EU MDR/IVDR Trained 4 International Projects
Download CV Contact Me
Houssem Eddine

About Me

EU MDR/IVDR-trained AI Engineer with production experience in clinical-grade medical imaging, computer vision, and agentic systems. Delivered results across medical, industrial, and geospatial AI — cardiac segmentation pipeline (Dice 0.88–0.91) validated by senior cardiologists, defect detection reducing manufacturing return rates from 6.8% to 0.2%.

Available immediately for remote or relocation-based roles.

🇬🇧 English C1 🇫🇷 French B2 🇸🇦 Arabic C1 🇩🇪 German B1

Tech Stack

Medical Imaging
MONAI 3D3D U-NetDICOM (pydicom)SimpleITKOHIF Viewer3D SlicerONNXGradCAM/XAI
Deep Learning
PyTorchTensorFlowLLM Fine-tuning (SFT+DPO/RLHF)Transfer LearningYOLO
Agentic AI
LangChainLangGraphCrewAIMCP ServerAgent-to-Agent (A2A)
Deployment & MLOps
DockerONNX RuntimeREST APIsGitLinux
Metrics & Standards
Dice ScoreHD95IoUEU MDR/IVDRGDPR-compliant
Core
PythonSQLNumPyPandasScikit-LearnMatplotlib

Work Experience

AI Engineer – Computer Vision (Contractor) · SBT Tunisia

Jan 2026 – Jun 2026 · Tunis, On-site

  • Real-time YOLOv8 quality control pipeline for automotive wiring harnesses
  • Reduced client return rate from 6.8% → 0.2% (French + Italian clients)
  • Mobile CV pipeline: cable routing, colour sequencing, configuration compliance
  • Integrating A2A + MCP server for automated defect reporting; architecture ISO 13485-transferable
YOLOv8PythonOpenCVA2AMCP Server

Medical AI Engineer (Intern) · Alight Consulting, Germany (Remote)

Jun – Sep 2025

  • Multi-class 3D cardiac segmentation (LV, RV, myocardium): Dice 0.88–0.91, HD95 3.0–5.0mm, 18% above baseline
  • Blinded validation: 2 senior cardiologists, 40 cases, 95%+ clinical acceptance rate
  • EU MDR/IVDR technical documentation + GradCAM explainability maps
  • Fine-tuned clinical LLM (SFT+DPO) on 200+ reports: 35% less hallucination, 30% faster reporting
MONAI 3DPyTorchLangChainSFT+DPODICOMSimpleITKONNX3D SlicerOHIF Viewer

Geospatial AI Engineer (Contractor) · CMA CGM, France (Remote)

Jan – Jun 2025

  • U-Net on multimodal satellite imagery: 92.4% IoU (8% above baseline), 70% less manual labelling
  • 31% improvement in generalisation to unseen desert terrain
  • CrewAI multi-agent system: reduced weekly analysis from ~10 hrs to ~20 min
U-NetPyTorchCrewAIDocker

Data Scientist · Newrest, Tunis-Carthage Airport

Jun – Sep 2023

  • 3-day supply forecast model for full airport catering operation
  • Reduced food waste by 55%, eliminated stock shortages within 3 months
PythonScikit-LearnProphetSQL

Founder & AI Engineer – Retail Intelligence · Self-employed

May 2023 – Present

  • Demand forecasting: MAPE 52% → 88% accuracy, 42% less unsold inventory
  • VLM pipeline for customer preference analysis → +20% sales conversion
  • Social media trend detection 10–12 days ahead of demand peaks
  • Repaid 24-month bank loan in 6 months post-deployment
PythonPyTorchProphetLangChainVLMOpenCVFacebook Graph API

Research & Projects

Research · Preprint

NeuroAI-Pipeline

Sep – Dec 2025 | ESPRIT University × Research Dept

  • Multimodal brain MRI analysis: VLMs + multi-agent LLMs for clinical reasoning in tumor analysis
  • MedViT + FAISS case retrieval + U-Net/ResNet segmentation + GradCAM
  • 3.2× processing throughput, sub-300ms inference
  • 3-LLM agentic architecture (Gemini + LLaMA 8B + LLaMA 70B) → structured PDF clinical reports
  • Virtual consultation pipeline: text/audio(Whisper)/video → 5-agent RAG → full case in <2 min
PyTorchMONAIMedViTFAISSMongoDBLangGraphGeminiWhisperRAG
GitHub → Preprint →

Production · EU MDR/IVDR

Cardiac AI Pipeline

Jun – Sep 2025 | Alight Consulting, Germany

  • Production-grade cardiac MRI/CT segmentation, CE-mark-ready
  • Dice 0.88–0.91, clinically validated by 2 senior cardiologists (blinded study)
  • EU MDR/IVDR technical documentation, GradCAM explainability
MONAI 3DPyTorchDICOMSimpleITKONNX
GitHub →

🏆 Award · EdTech

Eduverse AI — Hackathon Talents Tunisie FMIA 2025

Nov 2025 | Cité de la Culture, Tunisia

  • 1st Prize · 5,000 TND · EdTech category
  • Converts any lesson (PDF, video, web) into interactive RPG-style AI learning games
  • Competed against top university teams nationally
LinkedIn Post →

Production · Geospatial

Satellite Reforestation AI

Jan – Jun 2025 | CMA CGM, France

  • AI-driven satellite segmentation for environmental planning & reforestation site selection
  • 92.4% IoU on multimodal satellite imagery, 8% above baseline
  • Multi-agent automation: 10 hrs/week → 20 min/week
U-NetPyTorchCrewAIDockerSatellite Imagery

Education

🎓 Diplôme d'Ingénieur in Computer Science – AI Specialisation

ESPRIT University, Tunis, Tunisia

2022 – 2027 (Currently Year 4 of 5)

GPA: 15/20 (French grading) ≈ 3.7/4.0 · Honours standing

Community & Leadership

🌐 Tech Manager — AI & Healthcare Education

DeepFlow, Tunisia · Jun 2023 – Present

  • Designed and led 20+ AI/ML workshops focused on medical imaging and healthcare AI
  • Reached 200+ students across ESPRIT, INSAT, and SUP'COM universities

Get in Touch

📧 Email
houssemeddineds@gmail.com
💼 LinkedIn
linkedin.com/in/houssemeddineds
🐙 GitHub
github.com/HoussemDs