Two of my papers were accepted at NeurIPS 2024: efficient MoE inference and parametric dimensionality reduction.