Abstract: To address the shortcomings of traditional preparation techniques in Bi-SiNW-Mg₂Si composite thermoelectric materials, such as difficulties in high-precision forming, SiNW agglomeration, and Bi segregation, a novel laser additive manufacturing process was proposed, integrating four-wavelength laser energy synergistic coupling, CNC precision powder feeding, and a six-axis parallel scanning platform. This breakthrough overcomes the limitations of low control precision and poor adaptability in conventional processes. To enhance controllability, an AI-BP neural network model was constructed and optimized through training on measured datasets, enabling real-time precise regulation of SiNW doping concentration, Bi doping ratio, and crystal orientation. This effectively suppresses SiNW agglomeration and Bi segregation, ensuring the material's ideal crystalline morphology and uniform microstructure. Experimental results demonstrate that the thermoelectric figure of merit (ZT value) of this material reaches 1.36, a 23.6% improvement over traditional methods, providing technical support and theoretical reference for its engineering applications.