Cédric Caruzzo

• Deceptive Grounding: conceived and led a first-author study of entity-attribution failure in clinical RAG, from benchmark design and software through formal analysis, production validation, visualization, and writing.
• Medical-model post-training and evaluation: own training and evaluation pipelines for large medical language models, including benchmark construction, retrieval, knowledge-grounded reasoning, and capability-retention analysis.
• Self-distillation: led comparative evaluation of self-distillation fine-tuning for medical-model adaptation and its integration into the post-training workflow.
• Pathology foundation models: develop self-supervised Vision Transformers for whole-slide imaging, studying positional encoding, optimization, and transfer at scale.
• Distributed inference: built fault-tolerant Python and NCCL orchestration for clinical-scale whole-slide image collections.
• Research culture: lead recurring paper-study sessions and write long-form technical articles on post-training, reasoning models, and evaluation.

Public result: Deceptive Grounding was evaluated across 13 models and 740 production drug-disease pairs. Read the preprint.

• Developed CellPainTR, a Transformer with source-context tokens and staged self-supervised and contrastive training for batch-robust Cell Painting representations.
• Evaluated out-of-distribution generalization under dataset, batch, and feature shift, including transfer to an unseen dataset without target-dataset fine-tuning.
• Developed ChemPainTR, a dual-encoder framework for bidirectional translation between chemical structures and cellular phenotypes.
• Completed the thesis Development of Foundation Models in the Context of Cell Painting as a KAIST Global Presidential Scholar.

• Built machine-learning classifiers and interpretable feature-extraction methods for MALDI-TOF mass spectrometry.
• Supported Wolbachia detection, mosquito-species identification, and leptospirosis studies with public-health and hospital collaborators.
• Contributed data analysis to the Scientific Reports paper on Wolbachia detection.