Saiteja Jonnalagadda

Building AI on Generated Data Without Exposing Sensitive Information

Abstract:

Reliable AI depends on large volumes of realistic data, yet the most valuable data is often the very information an organization cannot share. This keynote is about resolving that tension. I will show how teams can build and test AI systems on realistic generated data that carries the same complexity and imperfections as real records, so models can be developed and evaluated without ever exposing sensitive information. Healthcare is one of the most demanding places to solve this, and it is where I draw my examples. The talk covers how to produce generated data that behaves like the real thing, how to match and link records across systems without revealing an identity, where generated data genuinely helps and where it can quietly mislead you, and what it takes to run all of this accurately and safely at scale. The theme throughout is simple. You should not have to choose between building capable AI and protecting the people behind the data.

Brief Profile:

Saiteja Jonnalagadda is a Senior Cloud Engineer and distributed-systems specialist whose work sits at the intersection of artificial intelligence, machine learning, and large-scale real-time data systems. He builds the foundation that makes AI trustworthy in production, engineering enterprise data to be accurate, reliable, and safe enough for machine learning models to act on at scale. His work spans generative AI for producing realistic data without exposing sensitive information, machine learning for anomaly detection and data-quality engineering, and drift monitoring that keeps models reliable as the underlying data changes, along with privacy-preserving techniques for linking records across distributed systems. He is a published researcher in applied AI, with contributions on generated data for record matching and privacy-preserving record linkage, and his broader focus is the safe operationalization of machine learning at scale in demanding, regulated environments.