AI is rapidly changing how data engineers write code, build pipelines, troubleshoot failures, and maintain complex data platforms. But getting real value from AI requires more than giving an agent access to a repository and asking it to generate code. Harness Engineering is about designing the environment around AI agents — providing the right context, tools, documentation, tests, observability, guardrails, and feedback loops so they can reliably understand, modify, validate, and operate data systems. In this talk, we will explore what Harness Engineering looks like in the context of modern Data Engineering. From data contracts, schemas, lineage, and pipeline metadata to automated testing, CI/CD, observability, MCP, and agent-friendly repositories, we will discuss how to make data platforms AI-operable by design. The goal is to move from AI as a coding assistant to AI as an engineering collaborator: capable of diagnosing pipeline failures, implementing transformations, validating data quality, navigating complex dependencies, and safely contributing to production-grade data systems.
With two decades of experience in Information Technology, and postgraduate studies at FGV and Unicamp, Felipe Pereira is an Entrepreneur and Speaker. As an Entrepreneur, he has founded Máquina de Dados, a consultancy company that helps businesses be more strategic and innovative using Analytical Intelligence and Artificial Intelligence. He is also a Speaker, having lectured at major innovation & technology events such as Google Datafest and Singularity University.
