In December 2024, Yu Wang developed Tensor Direct Storage for Connectors in the LMCache/LMCache repository, focusing on backend development and distributed systems using Python. This work refactored connectors to store tensors directly, bypassing traditional data serialization and reducing overhead in tensor-heavy pipelines. By introducing new connector types for both bytes and tensors and updating existing connectors to support these formats, Yu Wang streamlined the connector interface and improved data throughput. Setting the default remote_serde to None enabled seamless end-to-end tensor storage across connectors. The project addressed performance and scalability requirements, demonstrating depth in system design and data serialization.