Worked on embedding sharding and TBE integration within the pytorch/FBGEMM and pytorch/torchrec repositories, focusing on distributed systems and deep learning infrastructure. Developed features to enhance observability and reliability by introducing embedding table sharding metadata and improving shard context propagation in TBE construction, leveraging Python and PyTorch. Addressed stability in distributed caching by implementing deep copy isolation for sharding cache parameters, reducing cross-table mutation risks in multi-table workloads. Maintained strong traceability through explicit commit references and PR linkage, supporting future maintenance. The work demonstrated depth in distributed data engineering and contributed to more robust, maintainable embedding workflows in production environments.