In March 2025, Jan van de Bon expanded the MTIA embedding sharding options in the pytorch/torchrec repository, enhancing flexibility and scalability for large-scale embedding workloads. By introducing support for TABLE_COLUMN_WISE, ROW_WISE, TABLE_ROW_WISE, and GRID_SHARD sharding types within the embedding types module, Jan enabled more granular control over memory usage and performance tuning. The work, implemented in Python and leveraging PyTorch, addressed the need for broader deployment scenarios in distributed systems and machine learning pipelines. This contribution deepened the module’s capabilities, supporting more efficient production embeddings and facilitating adoption of MTIA across diverse, high-throughput environments.