Worked on enhancing batch processing capabilities in the ml-explore/mlx-lm repository by implementing batching for both MambaCache and ArraysCache. Leveraged Python to introduce batch generation support and added right-padding masking, enabling efficient handling of multiple requests in parallel. Updated existing models and comprehensive tests to validate the new batching features and ensure system robustness. Enabled server-side batching within the mlx-lm service, completing an end-to-end integration that improved throughput for multi-request workloads. Focused on model optimization and rigorous testing throughout the development process, delivering measurable performance improvements for machine learning applications that rely on efficient batch processing.