<p>Expanding productivity has been a priority for airlines. Aero-engine components are crucial parts of an aircraft, and Inconel 718 is the most widely used superalloy. Several aero-engine components require holes, and helical milling can be a feasible hole-making process for this difficult-to-cut material. Optimizing the material removal rate while meeting surface roughness requirements is a challenging task due to the uncertainty in roughness prediction. Several variables influence surface quality, including the type of lubri-cooling. This work proposes a framework for optimization with constraint learning applied to enhance the productivity of helical milling in Inconel 718. The roughness constraint is learned as a function of helical milling parameters and lubri-cooling type, including emulsion and minimum quantity lubrication. Seven machine learning regression approaches are tested to learn the constraints. The error of the learned constraints is estimated through cross-validation and incorporated into the models. Cubist was the best method to approximate the constraint, and support vector regression was tested for comparison. Optimization through metaheuristics is performed, and results with emulsion provided the best outcomes due to the difficulty of minimum quantity lubrication penetration at higher cutting speeds. The ant lion optimizer, dragonfly algorithm, and moth flame optimizer performed better in terms of convergence and computational performance. The main contribution of this paper in artificial intelligence is the application of machine learning to learn the process constraint while accounting for the error of the learned constraints, estimated through cross-validation. Distinct metaheuristics were considered for optimization to guarantee convergence. The proposed approach is compared with one using decision tree regression approximation solved as a nonlinear programming problem.</p>