Worked on the instadeepai/Mava repository to update the FF-IPPO Store Experience example, ensuring compatibility with the latest Mava changes. Addressed a bug by refactoring the learner to accurately handle episode metrics and implemented robust device state initialization and cross-device replication. Improved the creation of dummy transitions to ensure the correct observation shape was maintained throughout the workflow. The work leveraged Python and advanced libraries such as Flax and JAX, applying reinforcement learning and system adaptation techniques. This update enhanced the reliability and maintainability of the example, supporting ongoing development and adaptation to evolving upstream changes in the codebase.