Decentralized AI: Data, Governance, and Personalization at the Edge (with Katharine Jarmul & Joe Reis)

​AI doesn’t have to live in a central server farm. As the AI stack matures, more teams are questioning the default cloud-first model, especially when privacy, security, cost, or cross-org coordination are on the line.

​In this episode, we’re joined by Katharine Jarmul (privacy engineer, researcher, and author) and Joe Reis (data engineering educator and co-author of Fundamentals of Data Engineering) to explore what decentralized AI and decentralized data really mean in practice.

​This isn’t theory: we’ll dig into real-world examples of on-device inference, federated analytics, internal silos, cross-cloud systems, and policy enforcement at the data layer.

​We’ll discuss:

​🔍 What decentralized AI actually means and why it’s not just a training problem

​🧱 Patterns like federated learning, on-device inference, and distributed data governance

​🔐 Why some orgs choose to make things harder: privacy, compliance, mission alignment

​🧠 How user-controlled models and personalization shift the architectural defaults

​⚠️ The realities: divergence, coordination, infrastructure overhead, and how to manage it

​Whether you’re designing AI for regulated industries, scaling across silos, or building for offline and personalized use, this conversation will expand your sense of what’s possible (and necessary) in decentralized AI.