Hamza Cherkaoui

I develop theoretically grounded diffusion and flow-matching methods for reliable generative modeling under long-tailed data distributions, with an emphasis on rare-mode coverage, transfer, and reproducible large-scale evaluation.