Vendor Lock-In in the AI Era: Owning the Seams

Architecting and building data platforms has always involved navigating vendor lock-in, but in the AI era, the old goal of "zero lock-in" is an expensive myth. As organizations adopt generative AI, the threat hasn't vanished—it has simply shifted higher up the stack. ​Instead of monolithic databases, modern lock-in hides within metadata catalogs, embedding spaces, and AI orchestration frameworks. How do we leverage powerful cloud capabilities without permanently trapping our core intellectual property? ​This session explores the pragmatic trade-offs of the modern AI stack. We will discuss practical strategies to own the seams—like prompts, evaluations, and vector logic—while confidently renting the underlying compute.

Sathyaram Sannasi is a Google Developer Expert and a Data Engineer specializing in using Data and AI/ML for business outcomes primarily on Google Cloud Platform. As a freelance consultant, he partners with businesses to implement solutions that transform data into decisions and products.