Worked on the apache/tvm repository to improve the Relax TFLite frontend by addressing operator call issues in activation functions. Focused on correcting operator mappings for functions such as LEAKY_RELU, HARD_SWISH, ReLU_N1_to_1, and LOG_SOFTMAX, ensuring accurate integration with Relax IR. Enhanced code maintainability by standardizing operator naming and expanding unit test coverage, which validated the behavior of multiple activation functions and reduced the risk of regressions. Utilized Python and TensorFlow, applying unit testing methodologies to verify fixes. The work emphasized careful code refactoring and test-driven development, contributing to more robust TFLite frontend integration within the TVM project.