Xinkai Song

Institute of Computing Technology

Papers

1

Total Citations

17

H-Index

1

About

Xinkai Song is a leading researcher at the intersection of artificial intelligence and computer architecture, with a primary focus on enabling efficient on-device inference for large language models (LLMs). His most impactful work, "Cambricon-LLM: A Chiplet-Based Hybrid Architecture for On-Device Inference of 70B LLM" (2024), addresses the critical challenge of deploying advanced models like 70B-parameter LLMs on resource-constrained edge devices, including smartphones and robotics. This paper, which has garnered 17 citations in a short time, introduces a novel chiplet-based hybrid architecture that overcomes the extreme arithmetic intensity and single-batch computing constraints of edge environments. Song’s contributions are pivotal for advancing user data privacy and network resilience without sacrificing intelligent capabilities. His research is shaping the future of edge AI, making powerful LLMs accessible in real-world, privacy-sensitive applications.

Research Focus

Key Achievements

1
H-Index
1
Papers
17
Total Citations
17
Avg Citations/Paper
🏆 Most Cited Paper
Cambricon-LLM: A Chiplet-Based Hybrid Architecture for On-Device Inference of 70B LLM
17 citations · 2024
📈 Most Prolific Year: 2024 (1 Papers)
🤝 Key Collaborators: 14
🏛 Institutions: Institute of Computing Technology

Top Papers

  1. 1

Key Collaborators

Contact & Links

Available for collaboration
Content generated · 12 days ago