Piles are vital foundation systems that require careful design optimization. However, engineers often overlook soil uncertainties in optimizing pile designs. Moreover, exploring the entire design space for optimization remains a challenge due to the extensive computational time and complexity associated with numerical models. To address these challenges, this study introduces a novel multi-objective reliability-based design optimization (MO-RBDO) framework for pile foundations, explicitly incorporating soil spatial variability. Our novelty in this work is integrating soil spatial variability into an extended simplified model with calibration, yielding an efficient surrogate over complex numerical models that face challenges in exploring the entire design space in MO-RBDO. The proposed framework consists of three main components: (1) a calibrated simplified nonlinear load-settlement model capturing the interactions between piles and soil; (2) random field modeling to represent critical soil properties identified through sensitivity analysis; and (3) an MO-RBDO procedure that minimizes costs while maximizing pile reliability and design robustness. A case study demonstrated the framework's effectiveness, highlighting the impact of spatial variability on optimal pile design and the trade-offs among cost, reliability, and robustness. This approach offers engineers a more practical and economically sound design framework for pile foundations in heterogeneous ground conditions.