SentinelAura
Kubernetes-native visual arbitrage engine built on client-go and 3D WebGL. Continuously reconciles spot-instance pricing across 6 regions, automating node draining to the cheapest available zone.
The Problem
Static Kubernetes deployments ignored $140K/year in idle compute across over-provisioned, high-cost regions because operators had zero real-time visibility into cross-region spot-pricing volatility.
System Architecture
A Go controller using client-go watches EC2 pricing markets and reconciles pod placement via custom Kubernetes CRDs. A React-Three-Fiber 3D Earth renders live data-center heatmaps over WebSockets — turning cloud spend into a visual asset operators can act on in seconds.
System architecture diagram — coming soon
Technical Challenges & Trade-offs
Synchronizing cluster state across 6 regions without introducing latency overhead at the control plane. Implemented an eventual-consistency model over Redis Streams and offloaded heavy visualization math to Web Workers.
Business Impact & Metrics
Enabled real-time geographic cost arbitrage with automated node draining. Delivered 37% reduction in monthly compute spend and identified savings paths exceeding $420K/year in enterprise multi-region deployments.