Smart IoT systems require real-time and robust predictive analytics based on Machine Learning (ML) models. Building Big Data ML models is not only time consuming, but developers also lack the skills required for feature engineering, parameter tuning and model selection.The abundance of ML libraries and systems , data ingestion software, stream and batch processing engines, simulation technologies, and the number of hardware platforms accessible further exacerbates device architecture, fast development, and implementation issues. Finally, IoT's resource limitations demand that analytics engine execution be spread around the cloud edge spectrum. In order to address these challenging obstacles, we deliver Stratum, an event-driven Big Data-as-a-Service solution for IoT Lifecycle Analytics Management. Stratum provides developers with an elegant, declarative process, based on the model-driven architecture philosophy, to define device and network specifications. This paper illustrates the challenges Stratum is tackling, explaining its strengths through case studies in the modern world.