FNR / Industrial Projects

2026 – 2027 Luxembourg Institute of Health (PI: Olivier Keunen) FNR CORE project

Predi PET-AD aims to generate ‘PET-like’ images of molecular targets relevant to Alzheimer’s Disease (amyloid plaques, tau protein and metabolic activity) from multi-parametric MRI using AI, offering a way to reduce reliance on costly and hard-to-access PET imaging for diagnosis and clinical management. The project trains multi-targets vision transformer and diffusion models (ResViT, Brownian Bridge Diffusion) on large public cohorts (OASIS, ADNI, A4 — over 10,000 patients), validates the clinical value of the predicted imaging biomarkers, and investigates explainability techniques (Grad-CAM, LIME, SHAP) to support clinical adoption. I contribute my expertise in multi-targets AI models development.

2025 – 2027 Luxembourg Institute of Health (PI: Salah Ghamizi) FNR CORE project

FM2MRI aims to build the first foundation model for MRI modality synthesis and segmentation, combining Mixture of Experts and Retrieval Augmented Generation to tackle scarce medical imaging settings — rare pathologies and costly protocols such as DCE-MRI perfusion maps — where only limited training data is available. The project investigates unified brain MRI encoders, prompt-guided multi-task learning for joint segmentation and synthesis, and privacy-preserving customization through external database retrieval, aiming for a flexible, generalizable and locally deployable foundation model for clinical use.

2023 – 2025 FNR JUMP project: 250K€

SVALINN protects digital assets — like image, audio, video, text, and tabular data — from various misuses including secret disclosure, tampering, social engineering, copyright infringement and fake content generation. SVALINN relies on AI technologies to produce invisible signatures ensuring asset authenticity and impeding misuses by both humans and automated tools.

2020 – 2023 FNR CORE project: 910K€

In this project, we aim at complementing state-of-the-art machine-learning evaluation processes with testing techniques specifically adapted to the peculiarities of SLS. Indeed, although a plethora of techniques exists for testing traditional software, these are heavily challenged by SLS, their intrinsic probabilistic nature, their vast number of parameters, and their use cases too numerous to be elicited. More precisely, we focus on testing their underlying learning models and target three objectives: (1) measuring the adequacy of existing test cases with criteria that indicate how well the test cases cover the learning model; (2) defining model transformations (mutations) to modify the models, and estimating their sensitivity; (3) designing differential testing methods to discover disagreements between models, thereby obtaining new test cases that reveal errors in the models. Our three objectives are certainly not independent as fulfilling one will help achieve the others. Thus, altogether they will form a triangular chain of techniques to generate a high-quality test suite for learning models.

2020 – 2021 TruX Research Group FNR funded project: 120K€

Together with TruX Research Group, we built a simulation and policy recommendation tool for Covid19.

2019 – 2023 BGL-BNP Paribas

BGL-BNP Paribas is a leading bank of Luxembourg. We worked together to evaluate the robustness of their ML pipeline, in particular the themes of Concept Drift, Generalization and Adversarial Attacks.