In a boundary layer ingesting (BLI) propulsion system, the fan blades need to operate continuously under large-scale inflow distortion. The distortion will lead to serious aerodynamic losses in the fan, degrading the fan performance and the overall aerodynamic benefits of the aircraft. Therefore, in the preliminary design of a BLI propulsion system, it is necessary to evaluate the influence of the fuselage boundary layer under different flight conditions on the fan aerodynamic performance. However, a gap exists in the current computational methods for BLI fan performance evaluations. The full-annulus unsteady Reynolds-averaged Navier–Stokes (URANS) simulations can provide reliable predictions but are computationally expensive for design iterations. The low-order computational methods are cost-efficient but rely on the loss models for accurate prediction. The conventional empirical or physics-based loss models show notable limitations under complex distortion-induced off-design working conditions in a BLI fan, especially in the rotor tip region, compromising the reliability of the low-order computational methods. To balance the accuracy and cost of loss prediction, the paper proposes a data-driven tip flow loss prediction framework for a BLI fan. It employs a neural network to build a surrogate model to predict the tip flow loss at complex non-uniform aerodynamic conditions. Physical understandings of the flow features in the BLI fan are integrated into the data-driven modeling process, to further reduce the computational cost and improve the method’s applicability. The data-driven prediction method shows good accuracy in predicting the overall values and radial distributions of fan rotor tip flow loss under various BLI inflow distortion conditions. Not only does it have higher accuracy than the conventional physics-based loss models but also needs much less computational time than the full-annulus time-accurate simulations. The present work has demonstrated a significant potential of data-driven approaches in complex aerodynamic loss modeling and will contribute to future BLI fan design.