KubeCon + CloudNativeCon Japan 2025

As Generative AI reshapes the tech landscape, its tools and solutions often feel overwhelming and disconnected - like scattered elements waiting to be forged into something greater. What does a complete GenAI platform look like? And how does it differ from application-focused platforms many organizations already run?

We’ll guide you through a practical transmutation of your platforms to add GenAI features. We’ll first define the base elements needed to deploy LLM inference on Kubernetes. We’ll then blend key elements into the MVP stage, adding core enterprise AI components like an AI Gateway, shared interfaces via Dapr, and KServe-based deployment strategies. Finally, we’ll tackle advanced features - intelligent load balancing for LLMs, observability patterns, and cost optimizations.

More than a tools showcase, this session is packed with real lessons from enterprise GenAI platform engineers, to help transmute your platform into the philosopher’s stone for AI innovation.

Traditional service meshes often rely on sidecars, which can introduce added complexity and require application restarts. But things are changing. Modern service mesh is evolving into what we now call a service network—where the mesh dissolves into the infrastructure itself, becoming part of the infrastructure.

In this live demo, Lin will showcase how today's service mesh can be completely transparent to applications. She'll demonstrate how you can use gateways to implement Layer 7 traffic management and security policy enforcement, all while running her AI application that engages with and analyzes the audience's mood.

Migrating to Kubernetes offers numerous benefits, but few organizations start from "greenfields" circumstances. Many have critical services running on existing platforms, each leveraging unique capabilities. Transitioning these services to Kubernetes can be a complex and non-trivial challenge.

At Canva—a leading design and collaboration platform serving 220 million monthly active users—we recently undertook an ambitious project to migrate over 400 services from Amazon Elastic Container Service (ECS) to Kubernetes (EKS). These services now span 15,000 pods across two clusters on a typical day.

In this talk, we’ll share the strategies, tools, and processes Canva developed to achieve a (mostly) seamless migration with minimal disruption to our teams and users. Whether you're planning your own migration or just curious about what it takes to migrate at this scale, this session will provide practical insights and lessons learned.

Generative AI's rise demands massive compute resources and energy, conflicting with the urgent need for sustainability.

Composable Disaggregated Infrastructure (CDI) offers a solution by disaggregating servers into modular components connected via PCIe/CXL, enabling software-defined "composed server".

To address the growing resource demands of AI, particularly GPU utilization, Fujitsu is challenging the open-sourcing of a Kubernetes (K8s) mechanism for orchestrating CDI, under the CNCF community." This promotes vendor-neutral standardization, crucial for next-gen data centers handling explosive AI workload growth.

We aim to establish this as an open standard within CNCF/CNCJ, encouraging broad participation. Looking ahead, Fujitsu is developing "Optical Disaggregated Computing" for high-speed, low-power infrastructure in the 2030s.

Leveraging our CDI-based cloud technologies and photonics technologies, it delivers cloud-to-edge computing, improving efficiency, reducing power, and enabling low-latency.

Beyond K8s, Fujitsu's OSS contributions span AI, edge, and networking. We've strengthened our organization to focus on Linux/K8s and are committed to collaborating within the LF/CNCF ecosystem.

In the rapidly evolving world of cloud native technologies, new projects and paradigms are constantly emerging. To keep up with this momentum, platforms built on Kubernetes and cloud native ecosystem must continuously evolve as well. Choosing technologies and architectures is no longer a one-time decision—it requires continuous reevaluation and adaptability. By embracing flexibility and continuously adopting better solutions as they become available, end-users can maintain the competitive edge. This session will highlight the importance of continuously evolving platforms within the open-source ecosystem and introduce the powerful extensibility of Kubernetes that enables such evolution. Attendees will gain insights into how to keep competitive through continuous evolution, and we'll explore the next actions of early adopters and the trends in the upstream Just as the sun continues to rise in the Land of the Rising Sun, the evolution of your platform will illuminate the path to a brighter future for you, your company and your country.