I build systems that make neural networks smaller and faster.
Previously, I was a research engineer at Samsung Research.
International Conference on Machine Learning (ICML),
NanoQuant enables sub-1-bit large language models through post-training quantization (PTQ), using robust initialization and efficient reconstruction to outperform existing binary PTQ methods.
International Conference on Machine Learning (ICML),
RaBiT uses quantization-aware training (QAT), residual-aware optimization, and a dual-binary scheme to produce accurate, efficient binary large language models with performance comparable with state-of-the-art 2-bit quantization methods.