Shiyang Chen / The Mind's Tape
Black Square and Red Square (1915) by Kasimir Malevich
- I am a Ph.D. student at Rutgers University, focusing on High-Performance Computing. I am interested in the intersection of AI and HPC, particularly in the areas of large language models and graph neural networks. I will graduate at the end of 2025.
- I have experience in developing large-scale systems for AI workloads, including GNN inference and LLM inference.
- I have interned at several companies, including Amazon Web Services, ByteDance, and Microsoft, where I worked on various projects related to AI and HPC.
- I am passionate about computation efficiency and have a strong interest in exploring new architectures and algorithms to improve the performance of AI systems.
You can find my CV here.
news
| May, 2025 | I joined xAI as a Member of Technical Staff. |
|---|---|
| Mar, 2025 | My paper “Deal: Distributed End-to-End GNN Inference for All Nodes” is available on ArXiv. It proposes a set of systematic optimizations for end-to-end GNN offline inference. |
| Jan, 2025 | I will serve as a reviewer in MLSys 2025. |
| Jan, 2025 | I will serve as a reviewer in ICJNN 2025. |
| Dec, 2024 | My paper “KVDirect: Distributed Disaggregated LLM Inference” is available on ArXiv. It proposes a solution for the KV cache transfer in PD disaggregated LLM inference. It leverages the GPU RDMA to tackle the inefficiency and limitation in existing solutions such as NCCL and UCX. |
latest posts
| Mar 21, 2025 | Weight sharing in GPT |
|---|---|
| Feb 22, 2025 | Know your prompt template |