Verifiable Decentralized AI Pipelines using Zero Knowledge Proofs
In this talk we will discuss an approach based on cryptographic techniques like zero knowledge proofs to augment prevailing AI/ML practices to adequately address concerns around privacy and provenance. We are particularly interested in setting involving several participants such as data owners, data curators, model builders and finally the consumers of AI models, and how we can adequately address privacy of assets such as data and models, while simultaneously addressing concerns of end consumers regarding provenance of the results.
I am a researcher at IBM Research India, working in the area of applied cryptography with focus on privacy preserving machine learning, decentralized identity etc, employing cryptographic techniques such as zero knowledge proofs and multiparty computation. I completed my Bachelors and Masters in Computer Science from IIT Delhi and later obtained a PhD in Mathematical Sciences from Indian Institute of Science, Bangalore in 2014. In between, I have worked as a Software Engineer at VMware and Veritas.