Worked on the apache/tvm repository to deliver frontend support for the TFLite CUMSUM operator, enabling cumulative sum operations with axis and exclusive options in TVM Relax. The implementation involved mapping TFLite semantics to relax.op.cumsum, parsing axis values from constant tensors, and extracting CumsumOptions to ensure accurate parameter handling. Output data types were derived to align with model requirements, and groundwork was laid for future operator enhancements. This feature improves model compatibility and portability for edge and mobile deployments. The work was carried out using Python and leveraged skills in TensorFlow, data processing, and frontend development within machine learning pipelines.