KubeCon + CloudNativeCon India 2025 wrapped up in Hyderabad last week, bringing together thousands of developers, platform engineers, and cloud native practitioners for two days of intensive learning and collaboration. The second annual India edition showcased not just the technical evolution of Kubernetes, but the fundamental shift in how organizations think about developer experience, platform engineering, and AI workloads across cloud, data centers, and edge environments.
Akamai participated in the event as a gold sponsor, demonstrating product capabilities designed to simplify Kubernetes operations and support AI workloads at scale.
Below, I’ll walk through some of the key takeaways and technologies that emerged from the conference, including K8s as an operating system for Gen AI workloads that will shape cloud native strategies in the coming year.
AI and Kubernetes: From Experimentation to Production Scale
The convergence of AI/ML workloads with Kubernetes has progressed decisively from a proof of concept to a production reality. Intuit‘s keynote drove this home by highlighting its AI-native platform, which increased velocity by 8x for 8,000 developers. A keynote by Janakiram MSV showed why Kubernetes should be the operating system for Gen AI.
Throughout the conference, the message was clear: Kubernetes has become the de facto orchestration layer for AI workloads.
Key developments included:
- GenAI workload orchestration
Instead of relying on custom tooling, teams are increasingly using native Kubernetes primitives like CustomResourceDefinitions and StatefulSets. This means that AI/ML pipelines can be expressed in the same language developers already use for other workloads. - GPU scheduling optimization
Training and inference for large language models (LLMs) require efficient GPU allocation. New advancements allow Kubernetes schedulers to better match workloads to GPU resources. - Distributed AI agent coordination
Many AI applications are a collection of cooperating agents. The Kubernetes orchestration layer is now being leveraged to manage these distributed systems, ensuring they scale and communicate reliably. - Real-time AI inference at the edge
Instead of always sending data back to the cloud, inference can run locally on the edge. This avoids potential cloud latency and addresses performance without privacy trade-offs.
For organizations running AI workloads, managed platforms like Linode Kubernetes Engine (LKE) provide the foundation needed for these demanding applications. LKE supports GPU instances, autoscaling, and ML framework integrations, helping developers deploy inference services without the complexity of managing underlying infrastructure.
Kubernetes at the Edge: Solving Real-World Challenges
Edge computing emerged as a major theme, with multiple sessions demonstrating how lightweight Kubernetes distributions are enabling compute closer to data sources. These sessions also addressed challenges in edge computing, emphasizing that scale isn’t just about running Kubernetes on smaller devices, it’s about managing hundreds or thousands of distributed clusters efficiently.
Scale is something that Kubernetes is good at solving for, as it provides consistent orchestration across environments. When paired with AI workloads, edge computing is particularly powerful. Speakers from open source solutions like K0s demonstrated how real-time inference at the edge can eliminate cloud latency while respecting data sovereignty and privacy requirements.
Platform Engineering: The Answer to Kubernetes Complexity
Platform engineering has graduated from buzzword to critical discipline. The conference made it clear that raw Kubernetes is too complex for most developers to manage and maintain. They need abstractions that provide power without the pain.
Several speakers demonstrated how unified developer experiences can bring order to chaos and reduce onboarding time from weeks to days.
Akamai App Platform exemplifies these platform engineering principles by making Kubernetes production-ready out of the box. It eliminates the complexity of deploying and managing Kubernetes applications by providing a pre-configured stack of tools for CI/CD pipelines, network policies, storage, and observability, as well as golden path templates so developers can get the power of Kubernetes without the operational overhead.
eBPF and WebAssembly: Production-Ready Technologies
Two technologies that have been on the horizon for years finally demonstrated that they’re ready for production:
eBPF (Extended Berkeley Packet Filter) has become essential for:
- Non-invasive performance monitoring without application changes
- Kernel-level network security enforcement
- Zero-overhead observability for troubleshooting
- Real-time traffic analysis and filtering
WebAssembly (Wasm) is gaining traction in:
- Edge computing scenarios where containers are too heavy
- Edge and serverless functions.
- Plugin systems for extending platform capabilities
- Cross-platform portability without containerization
These technologies are particularly valuable for organizations using LKE, where eBPF can provide deep observability into cluster behavior and WebAssembly can enable lightweight workloads at the edge while maintaining integration with core Kubernetes infrastructure. Through our partner Fermyon, WebAssembly functions can run on Akamai, showing a new approach for fast, lightweight serverless workloads at the edge. Fermyon’s integration is perfect for event-driven architectures and low-latency use cases. Fermyon is also the creator of the CNCF projects SPIN and SpinKube.
Looking Ahead
The insights from KubeCon + CloudNativeCon India 2025 translate into clear actions for organizations:
- Evaluate platform engineering as a strategic initiative, not just a technical one
- Prepare for AI workloads by ensuring your Kubernetes infrastructure can handle GPU scheduling and distributed training
- Explore edge deployments for use cases requiring local processing, reduced latency, or data sovereignty
- Prepare for multi-cluster management to become essential as organizations distribute workloads across cloud, data centers, and the edge
- Understand that developer experience determines velocity as every friction point in your platform directly impacts business outcomes
- Consider managed solutions like Akamai App Platform and LKE when operational simplicity matters more than granular control
As multiple speakers emphasized, the future belongs to platforms that provide Kubernetes power without its pain, enabling developers to focus on what matters i.e., shipping code that delivers business value.
Based on this year’s momentum, next year’s KubeCon India in Mumbai promises to showcase even more innovation from India’s thriving cloud native ecosystem.
Comments