In modern industrial environments, ensuring the continuous and uninterrupted operation of machines is essential, as even a single motor failure can lead to substantial production losses and serious financial damage. Predictive maintenance focuses on identifying failures before they occur, but many existing industrial solutions are expensive, difficult to adapt, and lack the flexibility required for different applications. This project aims to create a modular, wireless, and affordable predictive maintenance system that can be easily integrated into production lines, educational environments, and the everyday operations of small businesses. The heart of the system is an ESP32 microcontroller paired with an MPU6050 gyroscope. This sensor measures motor vibrations along the X, Y, and Z axes, capturing detailed information about mechanical behavior. The collected vibration data is processed in batches of 1000 samples and then transmitted via WiFi to a FastAPI-based server. All measurements are stored in an SQLite database, allowing efficient handling and analysis. Before interpretation, the signals undergo Kalman filtering to reduce noise and improve accuracy. After filtering, the system applies a Fast Fourier Transform (FFT), which makes it possible to detect vibration peaks, frequency shifts, and atypical patterns that may indicate the early stages of mechanical failure. To ensure usability, a web interface was developed where users can monitor the status of each motor in real time. Several visualization modes are available, including last recorded values, tabular views, histograms, and scatter plots. In addition, an automated email notification system alerts users whenever unusual vibration behavior is detected compared to the motor’s normal operational pattern. For validation, a miniature demonstration model and a multi-motor test bench were constructed. During testing, motors with intentionally introduced mechanical faults produced vibration patterns that differed significantly from those of healthy motors. The system reliably identified these anomalies, demonstrating the effectiveness of the approach. The modular architecture also allows additional sensors—such as temperature or sound sensors to be integrated, and the server can be expanded to cloud-based operation, improving scalability and enabling broader industrial deployment. The completed prototype proves that it is possible to develop an effective, low-budget predictive maintenance tool.