Worked on ColossalAI and intel/sycl-tla, delivering features and fixes that improved model support, CI reliability, and documentation clarity. Enabled Qwen3 model sharding in ColossalAI’s ShardFormer by updating model policies, build pipelines, and test infrastructure using Python and PyTorch. Enhanced CI/CD stability through dependency pinning and workflow scheduling, and managed model zoo compatibility with PyTorch 2.5.1. In intel/sycl-tla, addressed a documentation typo and corrected a static assertion in CUDA code, updating TiledMMA configuration to support larger K dimensions. Demonstrated strengths in deep learning, distributed systems, and performance optimization, with a focus on maintainability and technical accuracy.