The New Stack Podcast

Deepak Goel, Nutanix Chief Technology Officer, explores the complexities of modern cloud environments. He highlights the struggle companies face in managing legacy systems while adopting microservices and Kubernetes. Goel discusses how platform engineering can streamline operations and the vital role of internal developer platforms in reducing friction. He unveils Nutanix's latest advancements in Data Services for Kubernetes, emphasizing enhanced security and data protection. Additionally, he shares insights on integrating AI to simplify coding and improve infrastructure efficiency.

Joining the conversation are Andrei Gueletii, a Google Cloud Technical Solutions Consultant; Pranay Bakre, a Principal Solutions Engineer at Arm; and Gari Singh, a Product Manager for Google Kubernetes Engine. They dive into why Google's Axion CPUs offer a compelling alternative to traditional GPUs, emphasizing cost-efficiency and energy savings. The guests discuss the flexibility of custom machine shapes in GKE, the shift to ARM architecture, and how many AI workloads can run more efficiently on CPUs. Insights into deployment at the edge and optimizing resource allocation add to the appeal.

Thanos Matzanas is an IBM observability expert focused on AI production, while Martin Fuentes is an AI ops practitioner specializing in product management. They delve into how organizational silos hinder AI deployment, emphasizing the importance of breaking down barriers between data scientists and operations. They discuss using OpenTelemetry for observability, the challenges of measuring AI performance, and the critical role of human feedback. Both guests stress that AI should enhance human expertise, particularly in high-stakes environments, while advocating for business-focused AI applications.

Rob Whiteley, CEO of Coder and veteran in open source and developer tooling, shares his insights on the rapid evolution of AI. He compares today’s AI developments to the Gold Rush era, emphasizing the critical role of tool providers. Whiteley highlights the skills gap and how it creates a divide among companies leveraging AI effectively. He introduces Coder's Mux, an innovative tool for managing multiple AI agents, and discusses the challenges of governance and agent accountability, emphasizing a new era of AI-native development.

Kevin Klues, a distinguished engineer at NVIDIA, and Jesse Butler, a principal product manager at AWS, explore cutting-edge advancements in Kubernetes for AI. They discuss the transformative Dynamic Resource Allocation (DRA), which allows users to specify GPU types and configurations, enhancing GPU usability. Klues highlights an upcoming workload abstraction aimed at optimizing complex AI workloads, ensuring coordinated multi-node jobs. They emphasize community involvement in shaping Kubernetes’ AI capabilities and the importance of balancing efficiency with system complexity.

Emilio Salvador, VP at GitLab, dives into the evolving role of developers at KubeCon North America 2025. He introduces the concept of 'cognitive architects,' who manage hybrid human-AI teams, and the 'AI guardian,' tasked with ensuring the quality of AI-generated code. Salvador discusses the AI paradox where coding speeds up but productivity stalls and emphasizes the need for AI integration throughout the development lifecycle. He also delves into the vision of a meta agent that acts as a proactive team member, advocating for gradual adoption and experimentation in organizations.

Jonathan Bryce, the new Executive Director of the Cloud Native Computing Foundation, shares his vision on the future of computing, emphasizing the crucial role of inference over training. He believes cloud-native infrastructure is ideal for scaling AI deployment efficiently. Topics include optimizing Kubernetes for managing GPU workloads and the launch of a Kubernetes AI Conformance Program. Bryce predicts that smaller, task-specific models will enhance performance as agents increase inference demands, laying the groundwork for the biggest workloads in history.

Dilpreet Bindra and Himanshu Singh, leaders at Broadcom/VMware, discuss the new Certified Kubernetes AI Conformance Program that aims to streamline AI tooling across Kubernetes platforms. They delve into the importance of portability and reliability to ensure seamless migration of AI workloads between clouds. The duo shares insights on VMware's integration with Kubernetes, highlighting how the vSphere Kubernetes Service (VKS) is among the first certified platforms, ultimately enhancing developer experience and enterprise trust in Kubernetes for production-grade AI.

Marek Szakowicz, lead maintainer of etcd and senior software engineer at Google, discusses tackling the critical challenges faced by the etcd project due to maintainer turnover and lost knowledge. He dives into the innovative approaches the team adopted, such as robustness testing inspired by Jepsen to ensure system correctness. Marek also details the collaboration with Antithesis to implement deterministic simulation testing, which enables reproducible outcomes and enhances reliability. His insights underscore the importance of rigorous testing in open source projects.

In this discussion, Matt Butcher, CEO of Fermion Technologies and co-creator of Helm, along with Matt Farina, Chief Architect at SUSE, delve into the exciting launch of Helm 4. They highlight Helm's evolution from a hackathon project to a key Kubernetes package manager and discuss the importance of WebAssembly-based plugins for cross-platform support. They also explore enhancements like modernized logging and the re-architecture of internals, emphasizing the value of 'boring' improvements that streamline workflows and future capabilities.