Running AI/ML workloads at the edge introduces challenges that traditional cloud-only architectures cannot solve latency, unreliable networking, heterogeneous hardware, and distributed coordination. This session shares practical lessons from building High performance system, a cloud-native, Kubernetes-aligned edge orchestration framework designed to run AI/ML inference and distributed compute across thousands of GPU/NPU/CPU edge nodes.

The talk covers:

1. Kubernetes patterns for deploying and managing edge AI workloads

2 Running ML pipelines across mixed hardware with containerized microservices

3. Real-time inference, scheduling, observability, and reliability at the edge

4. How we combined cloud-native DevOps + AI/ML + distributed systems to achieve resilient deployment

5. Practical failures, debugging war stories, and performance optimizations

6. Attendees will take away actionable patterns for building high-performance, production-ready edge AI systems using Kubernetes and CNCF tools.