This article presents an anomaly detection system for bearing data, leveraging Apache Kafka for efficient data streaming and an LSTM autoencoder. The system relies on a data producer that reads bearing vibration data, computes the Root Mean Square values, and streams both raw and processed data to a Kafka topic. The consumer processes this data, storing the raw values in a PostgreSQL database, and performs anomaly detection. By utilising pre-fitted scalers and defined statistical thresholds, the system effectively identifies deviations from normal operating conditions. The architecture ensures seamless data ingestion, storage, and processing, enabling timely interventions to prevent equipment failures. Experimental results demonstrate the system’s efficacy in detecting anomalies, highlighting its scalability, adaptability, and practical applicability in industrial predictive maintenance.