This chapter concludes the thesis by summarizing the core contents and contributions and providing possible future research directions. Specifically, it reviews the integration of Computer Aided Geometric Design (CAGD) with Deep Learning (DL) to develop robust adaptive fitting schemes using THB-splines. The conclusion highlights the main thesis contributions, such as enhanced fitting via iteratively reweighted least squares and quasi-interpolation, data-driven parameterization through (graph) convolutional neural networks, and the design and development of the "moving parameterization" paradigm within adaptive (THB-)spline schemes. Finally, it outlines critical future research directions: employing DL for automatic boundary detection, developing quasi-conformal parameterizations to minimize geometric distortion, and extending the proposed methodologies to multi-patch frameworks for industrial CAD design.