AI‐Driven Thermoelectric Generator Design: From Finite‐Element Simulation to Composable Device Assembly

The rapid expansion of thermoelectric materials has created unprecedented opportunities for waste-heat recovery, but it has also shifted the central challenge from discovering highperformance materials to efficiently assembling them into optimal device architectures.Thermoelectric generators convert temperature gradients directly into electricity, offering a solidstate route for low-grade waste-heat recovery and distributed energy harvesting [1,2].Recent advances have produced an increasingly diverse library of thermoelectric candidates, including Bi 2 Te 3 -based alloys, MgAgSb and Mg 3 Sb 2 compounds, and GeTe-and SnSe-based systems [3][4][5][6].However, translating this expanding materials landscape into efficient devices requires simultaneous optimization of material pairing, segmented structures, geometric parameters, and operating conditions, creating a rapidly growing design space.Each additional material introduces new possibilities for n-p pairing, segmented-leg construction, and temperature-range matching, which must be further optimized together with segmentation ratio, leg geometry, boundary temperatures, and operating current [4,7].Because these variables are strongly coupled, seemingly promising material combinations can still underperform when their transport characteristics, geometric dimensions, and operating conditions are poorly matched.The challenge, therefore, extends beyond selecting high-zT materials to efficiently navigating a multidimensional design space in which materials, geometry, and operating conditions must be optimized simultaneously.Finite-element modeling, such as COMSOL, can accurately resolve coupled thermal-electrical transport and remains a reliable tool for evaluating specific device configurations [3].However, exhaustive screening rapidly becomes computationally prohibitive, as every new configuration demands repeated geometry construction, meshing, boundary-condition specification, and solution of coupled