Resume

Institut für Neuro- und Bioinformatik, University of Lübeck

• Built an end-to-end biomedical NLP system for NER, UMLS entity linking, coreference resolution, and LLM-driven causal relation extraction, structuring outputs as a provenance-aware Neo4j knowledge graph.
• Fine-tuned open-source LLMs including Gemma, LLaMA, and Qwen with LoRA/QLoRA, improving NER and relation extraction F1 by 21% and adding post-processing pipelines for data quality and coreference resolution.
• Developed generative diffusion models for realistic iSCAT nanoparticle microscopy image synthesis, improving data augmentation for downstream segmentation.

Institut für Neuro- und Bioinformatik, University of Lübeck

• Developed end-to-end deep learning pipelines for nanoparticle segmentation and size estimation from iSCAT microscopy images in collaboration with Harvard University.
• Trained and evaluated SAM, Mask R-CNN, CellViT, and UNet, achieving 76% F1 and 81% Dice on the held-out test set.

Institut für Neuro- und Bioinformatik, University of Lübeck

• Built and deployed an RL-based image enhancement system that achieved state-of-the-art results among RL methods on MIT-Adobe FiveK, reaching 26.75 dB PSNR and 0.85 SSIM, with a live demo app on Hugging Face Spaces.

Institut für Technische Informatik, University of Lübeck

• Designed deep reinforcement learning algorithms for adaptive beamforming optimization in mmWave/THz MIMO systems and evaluated deep learning methods for modulation classification.
• Published Fast and Efficient Reinforcement Learning of Beam Codebooks in mmWave and THz MIMO Systems in IEEE Transactions on Communications.

University of Lübeck, Germany

École Nationale Supérieure d'Arts et Métiers, Casablanca