Title: Data Science Without Data: An industry Perspective
Abstract: Data science — the ability to process, analyze, visualize, and leverage information — begins with data. The recent wave of data privacy, minimization, and sovereignty regulations make a private approach to data science the need of the hour. In this talk, we question this basic requirement of data science — can we perform data science without data?
I will describe our system Devron, which takes the first step in building such a platform. Devron is built upon the essence of Federated Learning (FL), a technology pioneered by Google that enables distributed machine learning while improving data privacy. Devron enables every step of the data science pipeline in a privacy-preserving manner – exploratory data analysis (EDA), querying, iterative model training, and model serving. I will use one of these features to share an industry perspective on building such a privacy-focused platform.
Bio: Sameer is the head of privacy at Devron Corporation. Prior to that, he was a postdoc at the RISE Lab at University of California, Berkeley and received his PhD from Princeton University. His research interests are in applied cryptography and in the design of efficient privacy enhancing technologies (PETs), focusing on the intersection between systems, machine learning, and cryptography. He has worked on a number of emerging PETs — multi-party computation, homomorphic encryption, federated learning, differential privacy and has contributed open source software to the research community. Some of his work has already seen adoption in Industry and his protocols are integrated in initiatives at privacy focused start-ups and research labs. His work has received recognition from organizations such as Facebook and Qualcomm.
Host: Yixin Sun, assitant professor of electrical & computer engineering and computer science.
Organizer: Cong Shen, assistant professor of electrical and computer engineering, University of Virginia.