Engineering Scalable AI Pipelines: A Cloud-Native Approach for Intelligent Transactional Systems

The growth in the need for intelligent automation of transactional systems requires the design of scalable and robust AI pipelines. This paper introduces the cloud-native architecture for building Scalable AI pipelines that can withstand transaction, high-throughput workloads with low latencies and high availability. By taking advantage of micro services, container orchestration, and serverless computing, the proposed framework accommodates modular AI parts, such as the data ingestion, real-time pre-processing, model training, and inference. We show the effectiveness of the system in its application on the Kubernetes-integrated cloud platforms with stress on the high degree of elasticity, fault tolerance, and constant model integration. Scalability, execution time, and system responsiveness are significantly improved on benchmark transactional datasets after performance evaluations are conducted: a dramatic improvement from traditional monolithic AI workflows. This research provides the blueprint towards the development of intelligent and resilient and optimised for the cloud AI-driven transactional systems.