Meher BOULAABI

“Making AI transparent and trustworthy for healthcare”

📚 Academic Profiles

👤 Research Profile

Doctoral researcher specializing in interpretable deep learning for medical imaging, with focus on vision transformers and explainable AI. Published first-author papers at AIME 2025 and IEEE AICCSA 2024. Teaching experience: 63h of TD/TP in AI, machine learning, and computer vision, delivered in French and English at university level through an academic training project. Currently co-supervising 3 Master’s students in medical imaging and NLP research.

Research Focus

My current research focuses on developing Concept Bottleneck Models (CBMs) for interpretable medical imaging, with particular emphasis on creating clinician-verifiable reasoning systems that bridge high-performance deep learning with explainable clinical decision support.

Career Objective

I am actively seeking an ATER (Attaché Temporaire d’Enseignement et de Recherche) position to further develop my research in interpretability and explainable AI while contributing to academic teaching. I am highly motivated to teach courses in machine learning, deep learning, computer vision, and AI for healthcare, sharing my research experience with the next generation of students. An ATER position would allow me to strengthen my academic pedagogy and build a solid academic career in France within a research and teaching environment.

💻 Technical Skills

Programming Languages
Python, SQL, Bash/Shell Scripting, MATLAB

Deep Learning Frameworks
PyTorch (primary), TensorFlow/Keras, Hugging Face Transformers, Scikit-learn

Medical Image Analysis
OpenCV, Scikit-image, Pillow | Modalities: Fundus, OCT, X-ray, CT/MRI

ML Techniques
Transfer Learning, Attention Mechanisms, Hyperparameter Optimization, Data Augmentation

Model Architectures
CNNs (ResNet, EfficientNet, U-Net, DeepLabv3+), Transformers (ViT, Swin), LLMs (BERT, LLaMA, GPT)

Explainable AI (XAI)
Concept Bottleneck Models (CBM), Grad-CAM, SHAP, LIME, Attention Visualization

Data Science
Pandas, NumPy, Matplotlib, Seaborn, Plotly | Statistical Analysis, Exploratory Data Analysis (EDA)

Development & Collaboration
Git/GitHub, VS Code, Jupyter, Google Colab, TensorBoard, Streamlit, Notion, Excalidraw

High-Performance Computing
SLURM cluster management, Linux/Ubuntu administration, distributed training

Research Datasets
APTOS 2019, IDRiD, Messidor-2, EyePACS, ChestX-ray14, CheXpert, HAM10000, ISIC, CUB-200

🔬 Research Interests

Artificial Intelligence in Healthcare
Explainable AI (XAI) & Model Interpretability
Concept Bottleneck Models
Medical Image Analysis
Vision Transformers & Deep Learning
Large Language Models (LLMs)
Computer Vision

🎓 Education

Faculty of Sciences of Monastir, University of Monastir

PhD in Computer Science - Artificial Intelligence in Healthcare
September 2023 - Expected December 2026
Monastir, Tunisia

Thesis: Explainable and Interpretable Deep Learning Models for Medical Image Diagnosis
Advisors: Prof. Afef Kacem ECHI (LaTICE) and Prof. Zied Bouraoui (CNRS CRIL)
Research focus: Developing intelligent systems for ocular disease diagnosis using artificial intelligence and medical imaging techniques with emphasis on interpretability and clinical applicability

Higher National Engineering School of Tunis (ENSIT)

Master of Science (M2) in Computer Science
September 2022 - October 2023
Tunis, Tunisia

Specialization: Computer Science & Artificial Intelligence
Thesis: A Deep Learning System for Diabetic Retinopathy Lesion Segmentation and Disease Grading
Achieved state-of-the-art performance in diabetic retinopathy analysis using advanced Transformer and CNN architectures

Higher Institute of Applied Sciences and Technology of Kasserine

Master of Engineering (M1+M2) in Embedded Systems
September 2021 - March 2024
Kasserine, Tunisia

Thesis: Dashboard for System Fault Prediction and Classification
Achieved 99% accuracy in predicting industrial system faults through signal detection and fault classification
Extensive exploration of machine learning models including Random Forest, Decision Tree, SVM, and optimization techniques

🔍 Research Experience

Doctoral Researcher - Concept Bottleneck Models

CNRS CRIL UMR 8188, Artois University, France
September 2024 - Present

Developing an interpretable deep learning framework based on Concept Bottleneck Models (CBM) for medical diagnosis with clinician-verifiable reasoning (manuscript in preparation)
Fine-tuned large language models (LLaMA-2, Phi-2, BERT) using Hugging Face Transformers for concept extraction and medical text understanding
Deployed models on CRIL HPC cluster (SLURM) with distributed training across multi-GPU infrastructure
Implementing explainability techniques to create transparent AI systems that provide human-interpretable intermediate representations

Doctoral Researcher - Vision Transformers for Medical Imaging

CNRS CRIL UMR 8188, Artois University, France
September 2024 - March 2025

Published first-author paper at AIME 2025: “Enhancing Diabetic Retinopathy Classification with Swin Transformer and Shifted Window Attention”
Designed and validated a reproducible preprocessing pipeline (CLAHE adaptive contrast enhancement, automated circular ROI cropping, targeted augmentation) improving model generalization by 20% across diverse clinical datasets with varying acquisition protocols
Achieved state-of-the-art performance: 89.65% accuracy on APTOS 2019 (n=3,662 fundus images), 97.40% on IDRiD dataset, with AUC-ROC: 0.94-0.98, demonstrating cross-dataset robustness
Conducted comprehensive ablation studies to validate the contribution of each preprocessing step and architectural component

Doctoral Researcher - Semantic Segmentation

LaTICE Laboratory, ENSIT, Tunisia
January 2024 - Present

Published at IEEE AICCSA 2024: “Advanced Segmentation of Diabetic Retinopathy Lesions Using DeepLabv3+”
Implemented DeepLabv3+ with Atrous Spatial Pyramid Pooling (ASPP) for pixel-wise lesion segmentation: 99% accuracy (Dice: 0.97, IoU: 0.95) on IDRiD dataset
Designed domain-specific preprocessing (automated ROI extraction, CLAHE on LAB color space), achieving 92-97% lesion-level sensitivity for microaneurysms, exudates, hemorrhages, and soft exudates
Developed evaluation framework comparing multiple semantic segmentation architectures (U-Net, PSPNet, DeepLabv3+)

Research Intern - Machine Learning for Fault Diagnosis

Universitat Politècnica de Catalunya (UPC), Barcelona, Spain
October 2023 - January 2024

Designed and deployed a Streamlit dashboard for photovoltaic fault prediction, combining interactive data visualizations with fault onset detection using time-series analysis
Achieved 99% F1 score through systematic evaluation of ML algorithms (KNN, Decision Tree, SVM, Random Forest)
Implemented feature engineering and correlation reduction techniques to handle high-similarity signal data

📄 Publications

Enhancing Diabetic Retinopathy Classification with Swin Transformer and Shifted Window Attention

Boulaabi, M., et al. (2025)
Artificial Intelligence in Medicine Europe (AIME 2025), Springer Nature

Novel approach to diabetic retinopathy classification using Swin Transformer architecture with shifted window attention mechanism. Achieved state-of-the-art results on multiple benchmark datasets with comprehensive evaluation across APTOS 2019 and IDRiD datasets.

Advanced Segmentation of Diabetic Retinopathy Lesions Using DeepLabv3+

Boulaabi, M., et al. (2024)
Proceedings of IEEE/ACS International Conference on Computer Systems and Applications (AICCSA 2024)

Comprehensive study on lesion segmentation in diabetic retinopathy using DeepLabv3+ architecture with ASPP. Demonstrated exceptional performance with 99% accuracy on IDRiD dataset for multiple lesion types.

👥 Supervision & Mentoring

Master’s Thesis Co-Supervision (3 M2 Students)

University of Tunis, ENSIT
February 2025 - December 2026

Thesis 1: Automated Melanoma Detection using Deep CNNs

Research protocol design, ensemble architectures (ResNet, EfficientNet, ViT)
Grad-CAM explainability implementation on HAM10000 dataset (10k+ images)
Focus on interpretable diagnostic systems for dermatological applications

Thesis 2: Comparative Analysis of Architectures for DR Detection with XAI

CNN vs. Transformer comparison for diabetic retinopathy detection
Implementation of explainability methods (Grad-CAM++, LIME, SHAP)
Cross-dataset evaluation and robustness analysis

Thesis 3: Fine-Tuning LLMs for English-Arabic Medical Translation

Domain adaptation of multilingual models (mT5, mBART) with LoRA
Dataset creation and model deployment strategies
Collaboration with Union of Arab Scientific Research Councils

💼 Professional Experience

Digital Strategy Consultant - CMS and Meta Ads

Fiverr (Freelance)
February 2018 - August 2022

Managed digital infrastructure, e-commerce platforms, and data-driven marketing campaigns
Achieved Level 2 seller status with 150+ completed orders
Generated 300+ product orders through strategic targeting and optimized conversion funnels
Developed comprehensive growth strategies resulting in 30,000+ total followers and 10M+ impressions
Transferable skills: Project management, data analytics, client communication, A/B testing, performance optimization

🎯 Teaching Experience

Computer Science Instructor

Association Jeune Actif (Academic Training Project)
March 2024 - July 2024

Delivered comprehensive technical training programs combining theoretical foundations with practical coding exercises. Developed complete Jupyter notebooks, coding exercises, and evaluation rubrics.

Course Portfolio:

Advanced Python for Machine Learning (20h, 2 groups, M1 level)

End-to-end ML project development using VS Code and Google Colab
Covered NumPy, Pandas, Scikit-learn, OpenCV, and complete ML pipelines
Hands-on training in data preprocessing, model training, and evaluation

Computer Vision with Deep Learning (15h, 1 group, L3/M1 level)

CNNs architecture and implementation for image classification
Transfer learning techniques using pre-trained models
Practical applications using TensorFlow and PyTorch

Deep Learning Fundamentals (12h, 1 group, L3 level)

Neural networks theory: backpropagation and optimization algorithms
Implementation of fundamental architectures from scratch
Training techniques and hyperparameter tuning

Linux & System Administration (8h, 1 group, L1/L2 level)

Command line proficiency and shell scripting
Package management and system configuration
Development environment setup for data science

Data Analysis & Visualization (8h, 1 group, L2/L3 level)

Exploratory Data Analysis (EDA) methodologies
Statistical analysis and hypothesis testing
Visualization using Matplotlib and Seaborn

Teaching Summary:

Total Volume: 63h TD/TP delivered across 5 courses
Student Impact: 5 groups, 40+ students trained
Instruction Language: French with English technical materials
Format: Hands-on practical sessions with comprehensive supporting materials

📜 Certifications

View all credentials on LinkedIn

AI & Machine Learning

Building Transformer-Based NLP Applications (NVIDIA, In Progress)
Fine Tuning LLM with Hugging Face Transformers (Udemy, In Progress)
Introduction to Large Language Models (LinkedIn Learning, 2025)
Convolutional Neural Networks in TensorFlow (DeepLearning.AI, 2024)
AI for Medical Diagnosis (DeepLearning.AI, 2023)
Supervised Machine Learning: Regression and Classification (DeepLearning.AI, 2023)

Programming & Web Development

Python Programming Bootcamp from Basics to Advanced (Udemy, 2023)
Web Development, HTML5, HTML & CSS (Udemy, 2021)

Business & Leadership

Entrepreneurship, Social Innovation, Communication, Management, and Leadership (U.S. Embassy Tunis, 2017)
Design Thinking, Business Model Development & Pitch Presentation (GIZ GmbH, 2017)

Conference Participation

ACS/IEEE AICCSA 2024 (Presenter)
AIME 2025 (Presenter)

🌐 Languages

French: Professional working proficiency (actively improving toward C1)
Daily immersion at CNRS CRIL (lab meetings, research discussions), 43h technical teaching delivered in French, ongoing French language courses

English: Full professional proficiency
Published research papers, international conferences, technical documentation

Arabic: Native speaker

🏆 Awards & Volunteering

Elite Freelancer Participant - Tunisia Program

Uprodit
February 2023 - April 2023

Selected among 20 elite freelancers in Tunisia for intensive training in communication and leadership. Presented work under the patronage of Madame la Cheffe du Gouvernement Najla Bouden and engaged with influential government and business figures. This program fostered invaluable connections and aimed to shape the future of freelancing in Tunisia.

Club Project Manager

Young Leaders Entrepreneurs
January 2017 - January 2018

Project Manager of “Peace Lab Kasserine,” established in a disadvantaged region of Tunisia. Led initiatives to empower youth and promote peace, art, and theater. Implemented programs fostering communication, community involvement, and socio-economic development. Engaged over 100 young individuals in various activities including collecting gifts for cancer-stricken children and organizing internships.

Active Theatre Artist and Cultural Animator

Ministry of Culture of Tunisia
January 2010 - January 2015

Led and organized over 100 community projects, internships, and large-scale summer programs across Tunisia. Collaborated with the Ministry of Culture and Ministry of Education on cultural development initiatives. These experiences fostered a deep commitment to societal change and cultural development.

Last updated: February 2026