<p>Interface flaws and misaligned energy levels can impair perovskite solar cells’ (PSCs’) performance. By enhancing the quality of the interface, you can obtain stable and efficient PSCs. In this article, we have proposed organic dye-based surface modification as a potential mechanism to develop the efficiency and moisture stability of PSCs. In the developed mechanism, we incorporated the Gated Dual-Stream Convolutional Neural Network (GDSCNN) and Pufferfish Optimization Algorithm (POA) mechanism in the proposed approach. The primary aim of this study was to engineer the performance of PSCs by maximizing the power conversion efficiency (PCE) while minimizing surface defects, nonradiative recombination, and enhancing moisture stability of the devices. In this way, we used GDSCNN to forecast the power conversion efficiency while the POA was used to optimize solar energy conversion. The proposed mechanism has been applied and compared with traditional optimization mechanisms, which included Genetic Algorithm (GA), Bayesian Optimization (BO), and Convolutional Neural Network (CNN). As indicated in Table <InternalRef RefID="Tab5">V</InternalRef>, the proposed mechanism had a peak PCE of 21.6% as compared with BO 16.5%, GA 12.7%, and CNN 11.4%. The results showed significant improvement in photovoltaic performance, demonstrating potential for future work in next-generation PSCs.</p>