I work on AI infrastructure, domain-specific languages, and high-performance operator and runtime systems. My current focus is making deep-learning workloads easier to optimize and deploy across GPUs and diverse AI accelerators.
I contribute to several open-source projects in the InfiniTensor organization, including:
- NineToothed — a domain-specific language that uses tensor-oriented meta-programming to make high-performance GPU kernels easier to write.
- InfiniOps — a high-performance, cross-platform operator library supporting multiple GPU and AI-accelerator backends.
- InfiniRT — a runtime library providing a unified entry point for device management, memory operations, and other accelerator runtime capabilities.