M.Eng (Mines Saint-Étienne) | M.Sc. Candidate (Université PSL)
I am an engineer graduated from Mines Saint-Étienne and Université Paris Dauphine - PSL, currently specializing in AI for healthcare within the Health Sciences & Technologies (ST4H) Research Master's at Université PSL.
I am currently a Research Intern at the Genome Institute of Singapore (A*STAR), working on Prot2RNA, a Diffusion Language Model for Protein-Conditioned mRNA coding sequence generation.
Keywords: MIMIC-IV, Time-Aware Transformer, Interpretability
- Developed a Transformer-based architecture to predict sepsis onset from irregular clinical time-series.
- Achieved 0.84 AUROC, matching LSTM baselines while providing interpretable Attention Maps to visualize shifts in clinical importance (e.g., from routine care to critical biomarkers).
Keywords: Benchmarking, GNN vs Feature Engineering, ChemBERT A comprehensive evaluation of Deep Learning limitations on small chemical datasets (BACE, SIDER).
- Feature Superiority: Demonstrated why classic models with strong feature engineering often surpass complex architectures (SVR vs. MLP/GNN) on limited data.
- Transfer Learning Limits: Analyzed the necessity of fine-tuning ChemBERT for specific toxicity tasks.
- Optimization: Applied Bayesian Optimization to navigate chemical space effectively.
Keywords: Computational Pathology, Multiple Instance Learning (MIL)
- Implemented Weakly Supervised Learning (MIL) on Whole-Slide Images (WSI) to predict genetic mutations from histopathology slides.
