Drilling of carbon fiber reinforced polymer (CFRP) composites is an indispensible operation to produce holes for the assembly of components. This study optimizes the cutting parameters to reduce exit delamination and surface roughness, increase production rate in CFRP drilling. Firstly, a full factorial experiment is carried out to examine the effects of spindle speed and feed rate upon exit delamination and surface roughness of drilled holes. Analysis of variance (ANOVA) of experimental data indicates that feed rate has predominant influences on both delamination factor (Fd-out) and average surface roughness (Ra), accounting for large contributions of 93.74% and 70.39%, respectively. Secondly, a multi-objective optimization model is constructed with Fd-out, Ra, and material removal rate (MRR) as the optimization goals; Fd-out and Ra are related to the cutting parameters by regression analysis. Modified non-dominated sorting genetic algorithm II (NSGA-II) is applied to solve the optimization model. The obtained Pareto front consists of 195 Pareto optimal solutions widely distributed in the objective space, and the reliability of Pareto front is checked with the global convergence and spacing distance. Finally, preference relations-based decision-making is implemented to identify key solutions of better performance tradeoffs from the Pareto front. This study provides a feasible way to determine the appropriate cutting parameters, with which demands for multiple responses could be satisfied simultaneously in practical machining operations.