In this paper, wave tank experiments of a 1:10 scaled point absorber wave energy converter(WEC) are presented. The system consists of a six-degree-of-freedom ellipsoidal buoy connected via a lineto a linear eddy current brake power take-off (PTO) that emulates a damping force. The damping coefficientis adjusted by varying the number of neodymium magnets and determined through a series of drop tests. Thepower take-off’s position in various irregular sea states, generated from Bretschneider and JONSWAP spectra,is used to identify a deep neural network (DNN) based surrogate wave-to-motion model. The selected model notonly significantly outperforms the physical-based linear potential flow (LPF) model across all considered seastates, but also demonstrates encouraging generalization potential. The developed power take-off is consistentand provides suitable damping coefficients, at least for shorter waves, while also allowing clear pathways forimprovement to address a broader range of sea states.