Data-driven technologies, like Internet-of-Everything, which enables intelligence, rely on photoelectric elements, electronics, and communications. The conventional framework possesses optics, signal-processing, power supplies, and redundancies, which are riddles in its utilization. Optical wireless communication (OWC) can liberate such constraints, empowering widespread deployment to assemble information from surrounding objects and their energy efficiency. This study proposes a transparent photovoltaic (TPV) window that generates onsite-power and is suitable for indoor illumination for machine-learning. A TPV framework exhibited bifacial performance, including fast photoresponse and spectral responsivity, suitable for indoor illumination. For OWC, the spike-induced photoresponse is examined for pyroelectric ZnO and silver nanowire-embedded TPV devices, resulting in a notable binarization scheme and superior reliability of the optical signal. Owing to an ultrafast spike in the current, which flips from negative to positive current values and vice-versa upon optical signaling, preprocessing-free decoding of Morse-code is enabled to ensure the data-reliability of the optical signal of the 420 nm wavelength. In contrast to the Si device, the TPV device decodes a fast Morse-code carrying an optical signal >10,000 Hz with greater reliability. The results obtained for the TPV device with machine learning features provide a guiding significance for the development of high-performance, and reliable informatics.