Research interests are compiler optimizations for AI models with low-power numerics, especially in quantization and sparse optimization. He is good with LLVM, OpenCL, Halide, simulator, and TVM. He is with research experience in developing efficient scheduling primitives for sparse Halide programs with deep learning workloads and building OpenCL DSP feature set flow on LLVM compiler and GPGPU-Sim simulator. As a manager at PeakHills Group Corporation, he also supervises projects in groups.