The next generation of portable and wearable electronics demands the development of flexible and efficient energy storage systems. Supercapacitors are best suited for portable energy storage because of their high power density, fast charge/discharge capability, long cycle life, better safety, and lightweight properties. Extrusion-based additive manufacturing has recently gained attention to fabricate energy storage devices as it has better control over electrode design, layer thickness, and porosity, which enhances electrolyte ion diffusion and mechanical flexibility. In the extrusion-based direct ink writing (DIW) technique, a viscoelastic ink is extruded through a nozzle to build structures layer by layer. Manganese dioxide (MnO2) has attracted much attention as a supercapacitor electrode material because of its high theoretical capacitance, low cost, and environmental safety. However, selecting a proper method for fabricating supercapacitors, which ensures high electrochemical performance and flexibility requirements for wearables, is challenging. This work addresses such challenges by using 3D printing in combination with machine learning (ML) techniques to optimize the architecture and performance prediction of MnO2-based supercapacitors. This study used a DIW 3D printer (Hyrel Engine Hr) to fabricate customizable MnO2 electrodes for 3D supercapacitors. The optimization of the electrode parameters, such as electrode thickness and electrode area, is often very time-consuming and resource-intensive in a trial-and-error process. To address this issue, ML models were trained upon a database of electrochemical performance metrics and structural parameters. The ML model predicted critical performance indicators like specific capacitance, energy density, and cycle stability, significantly reducing the experimental iterations. The 3D printing of electrodes validated the ML-guided design under predicted optimal conditions. The electrode ink was prepared by mixing active material (MnO2), conductive agent (activated carbon) and binder (Poly(vinylidene fluoride)) with mass ratio of 8:1.5:0.5. The rheology and material characterization of the inks were performed using a rheometer, Scanning Electron Microscopy (SEM), X-ray diffraction (XRD) spectroscopy, Raman spectroscopy, and X-ray photoelectron spectroscopy (XPS). The electrotechnical characterizations, such as cyclic voltammetry (CV), galvanostatic charge/discharge (GCD), and electrochemical impedance spectroscopy (EIS), were conducted using a potentiostat/galvanostat (Gamry 1010E). The optimized printed supercapacitors exhibited a high specific capacitance of 7.65 mF/cm2, a high energy density of 0.27 mWh/cm2, and a power density of 13.18 mW/cm2 at a current density of 1 mA/cm2 and high stability over an extended number of charge-discharge cycles. In addition, the electrodes demonstrated great flexibility and mechanical robustness, with stable performance over different bending and other mechanical deformation conditions. The accuracy of the ML model was also cross-validated with the experimental data, confirming strong predictive capability over performance metrics of different architectures of MnO2 electrodes. This work highlights the synergy between ML and additive manufacturing for the rapid design and optimization of supercapacitors. Combining 3D printing and ML can significantly benefit the manufacturing of high-performance 3D supercapacitors for the next generation of flexible/wearable electronics.