This repository provides a Keras-based implementation of the Multi-gate Mixture-of-Experts (MMoE) architecture, a deep learning model that learns relationships between multiple tasks by leveraging shared expert networks with task-specific gating mechanisms. The implementation includes example demonstrations using the UCI census-income dataset and is designed to be extensible for other multi-task learning scenarios.