Modern applications increasingly demand lightweight, high-performance metallic materials, such as magnesium alloys, capable of exhibiting specific stress-strain responses under complex loading conditions. Achieving this requires precise engineering of the material's microstructure and texture at the meso-scale. While computational methods like crystal plasticity can predict mechanical properties from structure, their significant computational cost creates a bottleneck, particularly for inverse design where efficient optimization of microstructure for target properties is paramount. This necessitates an AI based surrogate model based on experimental data. In our research a comprehensive machine learning pipeline is developed to establish quantitative structure-property linkages in extruded magnesium alloys, using microstructure and texture descriptors. Leveraging experimental data from optical microscopy and X-ray diffraction, our workflow integrates automated image processing, deep learning for feature extraction, computation of advanced microstructure and texture descriptors (including gram matrices and generalized spherical harmonics), dimensionality reduction, and non-linear regression models (e.g., XGBoost, Gaussian Processes). The pipeline demonstrate the ability to accurately predict key mechanical properties like yield strength (σy) and strain hardening exponent (n) with high fidelity (e.g., MAPE ~7% using XGBoost) for unseen material specimen. Crucially, by employing interpretable AI techniques like SHAP (Shapley Additive exPlanations), we quantify the influence of specific microstructural and textural features (such as texture components and grain aspect ratio) on mechanical behavior. This derived feature importance provides actionable insights and serves as a critical input for inverse design. We discuss how this quantitative importance can establish a priority index, guiding efficient exploration of the design space using optimization algorithms like Bayesian Optimization, focusing the search on the most impactful features first. The proposed inverse design AI-driven framework significantly accelerates the discovery and optimization cycle for advanced magnesium alloys with tailored performance.