KubeWise, AI-Driven Kubernetes SRE Platform
Go, Kubernetes, Prometheus, Pydantic-AI, Helm
- Cut Mean Time To Detection (MTTD) by 60% and Mean Time To Resolution (MTTR) by 45% with an in-cluster agent that parses Alertmanager webhooks via Pydantic-AI and executes LLM-planned remediation: pod auto-scaling, node cordoning, traffic drain.
- Implemented 8 deterministic rules for OOM, CrashLoopBackOff, and node pressure, multi-strategy anomaly detection (Z-score, Bayesian changepoint, ROC scoring), and a 4-tier remediation policy with approval gates.
- Rolled out a Helm chart with observability subcharts and a tool plugin system covering kubectl, Helm, ArgoCD, Terraform, and GitHub CLI.