High-Flyer Quants Yinghuo II is a deep learning training platform that offers elastic, time-sliced shared compute with second-level scheduling response. It includes the hfainn high-performance operator library, the hfreduce distributed communication framework, and a purpose-built parallel file system (FS). Designed to maximize GPU utilization and scale models across nodes, it enables researchers and quant teams to train large models efficiently without dedicated hardware.