"From Problem to Production: DevOps for AI with Kubeflow" by Fabrizio Lazzaretti & Marco Crisafulli

Despite massive investments in AI, most enterprise AI initiatives fail to reach production — not because of model quality, but because of fragmentation among data science, platform engineering, and business teams. MLOps addresses this challenge by applying DevOps principles to machine learning, and Kubeflow has emerged as a de facto open platform for operationalizing these practices in cloud-native environments. In this talk, we explore how Kubeflow enables collaboration, automation, and repeatability across the entire ML lifecycle, from requirements gathering and experimentation to deployment and continuous improvement. Using a real end-to-end example, we’ll demonstrate how Kubeflow Pipelines, training workflows, and integration with the CNCF ecosystem help teams move beyond ad-hoc experimentation toward production-grade AI systems. Attendees will see how a vendor-neutral Kubeflow-based MLOps architecture supports multi-cloud deployments, enforces operational best practices, and creates continuous feedback loops between stakeholders. Whether you are introducing MLOps for the first time or refining an existing platform, this session provides practical insights into using Kubeflow to turn AI from a research activity into a sustainable business capability. Fabrizio Lazzaretti is a Managing Consultant at Wavestone and CNCF Ambassador who bridges cutting-edge cloud-native technologies with enterprise architecture. As maintainer of the CloudEvents Rust SDK and co-author of "Crafting Great APIs with Domain-Driven Design," he brings deep expertise in event-driven architecture and microservices to complex challenges. With over 10 years of experience in software architecture, development, and DevOps, he currently drives architectural transformation and AI initiatives across sectors, connecting business and IT through collaborative API design. Marco Crisafulli is the co-founder of enki, a Swiss machine learning company focused on training, consulting, and implementing practical ML use cases for Swiss enterprises. With a strong background in enterprise software and cloud-native infrastructure, he specializes in MLOps, Kubernetes, and production-grade ML platforms, and has deep hands-on experience with Kubeflow. He works closely with organizations to bridge the gap between experimentation and reliable ML systems in production, helping teams operationalize ML at scale.