The computational demands of physics-based simulations for lithium-ion battery (LIB) manufacturing present significant challenges for real-time control applications [2]. This research addresses the dimensional complexity of electrode coating processes — responsible for approximately 30% of production scrap — through novel surrogate-based optimization tailored for multi-physics systems with high parametric sensitivity. By integrating machine learning algorithms with high-fidelity multi-physics simulations, a parameter optimization framework is developed to achieve a desired coating profile, enabling therefore the possibility to minimize the defects present in a battery coating process [1]. The high-fidelity models, developed in Simcenter STAR-CCM+, serve as the basis for our surrogate framework and provide the training data for our hybrid modelling approach. The present modular surrogate approach preserves physical interpretability by separately treating the flow through the slot-die and its consecutive coating process. This decomposition framework serves as the foundation for a gradient-based optimization algorithm, which iteratively determines optimal T-Bar positional parameters to achieve specified coating thickness distributions. Such surrogate-based optimization workflow considers the essential dynamics of the coating process while achieving the computational efficiency required for industrial edge deployment. Overall, this contribution advances digital twin enabled process control by combining sensor data, CFD derived high-fidelity simulations, machine learning based surrogates, and gradient based optimization. The resulting system provides an efficient and physically grounded approach for improving coating uniformity and reducing scrap in LIB electrode manufacturing.