Nitrate (NO₃⁻) concentrations in aquifers constitute a global problem affecting environmental integrity and public health. Unfortunately, deploying hardware sensors specifically for NO₃⁻ measurements can be expensive, thereby, limiting scalability. This research explores the integration of soft sensors with data streams through an use case to predict nitrate NO₃⁻ levels in real time. To achieve this objective, a methodology based on Kafka-ML is proposed, a framework designed to manage the pipeline of machine learning models using data streams. The study evaluates the effectiveness of this methodology by applying it to a real-world scenario, including the integration of low-cost sensor devices. Additionally, Kafka-ML is extended by integrating MQTT and other IoT data protocols. The methodology benefits include rapid development, enhanced control, and visualisation of soft sensors. By seamlessly integrating IoT and data analytics, the approach promotes the adoption of cost-effective solutions for managing NO₃⁻ pollution and improving sustainable water resource monitoring.