Qiyu Lei developed MUSA backend support for the FlagEmbedding library in the upstash/FlagEmbedding repository, focusing on expanding device compatibility for enterprise machine learning workflows. By updating the device selection logic in both AbsEmbedder and AbsReranker, Qiyu enabled seamless integration of MUSA as a viable runtime option, allowing optimized embeddings on supported hardware. The work was implemented using Python and PyTorch, leveraging backend development skills to address deployment flexibility for clients relying on MUSA infrastructure. This targeted feature update improved the library’s adaptability, though the scope was limited to a single feature addition without bug fixes during the one-month period.