During his recent work, Wang contributed to both the ColossalAI and intel/sycl-tla repositories, focusing on deep learning infrastructure and documentation quality. He implemented Qwen3 model support in ColossalAI’s ShardFormer, enabling model sharding and updating the build pipeline for compatibility. To improve CI reliability, he pinned dependencies and adjusted workflow schedules, addressing flaky builds. In the model zoo, he managed compatibility by disabling unsupported models. For intel/sycl-tla, Wang fixed a correctness issue in the SGEMM SM80 example and enhanced documentation clarity. His work leveraged Python, CUDA, and YAML, demonstrating depth in distributed systems, dependency management, and performance optimization.