During September 2025, Fate developed fused Mixture-of-Experts (MoE) support for SM120 hardware in the flashinfer repository, focusing on hardware-optimized inference. Using C++, CUDA, and GPU computing techniques, Fate introduced the gen_cutlass_fused_moe_sm120_module to enable efficient MoE operations on SM120, while updating the SM100 build context to prevent cross-hardware conflicts. Fate also addressed stability by fixing FP4 MoE execution on SM120, ensuring reliable performance. These contributions improved MoE workload efficiency and reliability on modern hardware, aligned with customer deployment needs, and reduced maintenance risk, demonstrating depth in performance optimization and careful attention to hardware compatibility and deployment readiness.