The End of Manual DataOps: Git for Data and the Automation of AI Infrastructure

One area where we believe AI agents will be crucial is DevOps automation. Data and ML workloads have been historically very hard to automate reliably because of their inherent complexity, and because traditional data platforms require specialized expertise and manual oversight. This talk argues that the future of DataOps is built around code-first abstractions and fully-programmable data platforms that meet the needs of DevOps teams: full replicability, easy automation and transparent security. We will present very specific real-world examples on how Recommender and RAG systems running in production can become more reliable by abstracting infrastructure complexities behind familiar Python concepts like functions, packages, tables, and version control.