Abstract: In the field of quantum sensing, the design of micronano optical devices in graduate education is often hindered by interdisciplinary barriers and difficulties in bridging theory and practice. To address these challenges, this paper proposes an AIenabled, projectdriven training model that integrates “designsimulationoptimization” in a closed loop. The model orients students toward real research problems and guides them in applying AI algorithms to the design and optimization of micronano optical devices. A case study on deep learningbased phase modulation metasurface lens design illustrates how students progressively employ deep learning for device design and optimization. Results show that this approach enhances interdisciplinary competence, systems thinking, and research efficiency, thereby offering a pathway for cultivating highlevel talent in quantum technologies in the era of digital intelligence.

Key words: quantum sensing, artificial intelligence, micronano optical device design, innovation in teaching models