Data pipelines are often the most brittle part of our infrastructure, yet they are rarely tested for real-world failure. This talk introduces a production-proven SRE practice, Chaos Engineering, as a new standard for data quality and pipeline reliability. We will move beyond traditional data validation and explore how to proactively inject failures into our data stack. This session provides a practical framework for designing and executing controlled chaos experiments against critical data infrastructure, including ETL/ELT jobs, streaming processors, and data stores. We will cover how to simulate failures like network partitions, sudden data corruption, and resource exhaustion in orchestration systems. By discovering these hidden bottlenecks and single points of failure in a controlled setting, we can build truly resilient, production-grade data platforms. This talk is for engineers who want to move from a reactive to a proactive approach to data infrastructure.