Jiajun Kang

Efficient LLM inference, KV cache optimization, quantization, speculative decoding, model pruning