Worked on the pytorch/tensordict repository to deliver a flexible input-to-argument mapping feature within TensorDictModule. Introduced the out_to_in_map capability, allowing a single input feature to be reused across multiple arguments, which streamlines module configuration and reduces duplication. The implementation included deprecation warnings for legacy usage, guiding users toward the updated interface. Comprehensive tests were developed and executed to ensure the new mapping behaved as intended, and documentation was improved to support onboarding and reliability. This work leveraged Python, API design, and software engineering skills, enhancing modularity and test coverage in preparation for broader adoption within the codebase.