Anakin Lancer enhanced deep learning infrastructure across the FlagOpen/FlagGems and FlagTree/flagtree repositories, focusing on performance and hardware compatibility. He refactored the max_pool2d_backward_kernel in C++ to support an additional parameter and improved device handling, optimizing both forward and backward operations for diverse GPU environments. In Python, he expanded Pytest-based attention test coverage for Kunlunxin hardware, ensuring robust validation across builds. Anakin also introduced KunlunX SDNN backend support for Triton, integrating new analysis utilities and optimizations to streamline deep learning model execution. His work demonstrated depth in backend development, compiler design, and cross-platform machine learning system integration.