This work presents a methodology for optimizing a small-sized deep-drawn component using Sensitivity Analysis (SA) and Machine Learning (ML)-based optimization. The proposed approach aims to mitigate critical defects by controlling draw-in through the adjustment of key process parameters. The metamodel incorporates material properties (e.g., tensile strength, anisotropy), shim height, sheet position, sheet thickness, and friction. The ML model and sensitivity analysis were combined with an optimization framework. The optimization controller minimizes the deviation between the target draw-in—associated with defect-free parts—and the measured draw-in. The controller's performance was evaluated in various virtual scenarios where uncontrollable variables were altered. Results demonstrate that the optimizer effectively adjusted process parameters to achieve the target draw-in, ensuring robustness against process variations. This approach enhances process stability and defect prevention in stamping operations and is adaptable to other forming processes.