About me

I am a Research Software Engineer at Google Research, Australia, based in Sydney. I am currently focusing on ML efficiency, especially LLM inference optimizations. I obtained my Ph.D. from the School of Computer Science, Faculty of Engineering at The University of Sydney (USYD). Prior to that, I received my Master’s and Bachelor’s degree from the University of Science and Technology of China (USTC).

My research lies at the intersection of Computer Systems, Machine Learning (MLSys), and Hardware Architecture. I am passionate about breaking the memory wall and communication bottlenecks in generative AI. Through full-stack innovations across the algorithm, runtime, and hardware boundaries, my work primarily focuses on algorithm-system-architecture co-design for Large Language Models (LLMs), including:

• High-Performance Kernel Optimization: Designing efficient system acceleration and custom GPU/TPU kernels for Generative AI, including the kernel design of efficient MLP, Attention, and MoE layers.
• Extreme Model Compression: Developing low-bit weight/KV cache quantization (e.g., FP6, sub-byte formats) and unstructured sparsity exploitation to maximize inference throughput.
• Hardware Acceleration & Custom Microarchitecture: Leveraging domain-specific architectures (DSA) and custom hardware accelerators (e.g., FPGA/SoC prototypes) to co-design and execute hardware-native execution primitives for modern ML and pattern-matching workloads.

With systematic optimizations spanning from gate-level hardware logic up to distributed software systems, my ultimate goal is to build memory-efficient, high-throughput, and scalable infrastructure to democratize next-generation AI foundation models.