The rapid proliferation of Internet of Things (IoT) devices and smart city infrastructures has intensified the demand for real-time analytics capable of supporting intelligent services such as traffic optimization, energy management, public safety, and healthcare monitoring. Traditional cloud-centric data pipelines, while powerful, struggle to meet stringent latency, bandwidth, and privacy requirements inherent in these contexts. To address these challenges, edge computing offers a decentralized paradigm that brings computation and storage closer to data sources, thereby reducing transmission delays and alleviating pressure on centralized systems. This study investigates the design of decentralized data engineering pipelines that leverage edge computing to enable low-latency analytics in smart cities and IoT ecosystems. The proposed framework integrates data preprocessing, stream ingestion, and lightweight machine learning inference at the edge, while coordinating with cloud infrastructures for long-term storage, historical analytics, and large-scale model retraining. Key architectural considerations include distributed orchestration, workload partitioning, adaptive caching, and security protocols tailored for heterogeneous edge environments. Through simulated use cases such as real-time traffic monitoring and predictive maintenance of urban assets, the framework demonstrates reduced latency, improved fault tolerance, and enhanced scalability compared to purely cloud-based approaches. Additionally, decentralized architectures foster greater resilience by localizing critical decision-making, thereby ensuring continuity during connectivity disruptions. By uniting edge computing and decentralized pipeline design, this approach provides practical guidelines for building robust, low-latency analytics infrastructures. The findings underscore the transformative potential of edge-enabled pipelines in advancing smart city services and IoT applications while addressing critical requirements of timeliness, reliability, and efficiency.