Hi, I’m Andrey — a Platform Engineer and AI Infrastructure Architect building an open-source AI Infrastructure OS for governed private AI on Kubernetes.

I combine Kubernetes, GitOps, Infrastructure as Code, Observability, runtime engineering, identity, policy, FinOps and AI governance to build secure, scalable and governable AI platforms.

🔧 I design control-plane and execution-plane platforms with Kubernetes, OpenTelemetry, KServe, vLLM, KEDA, Argo CD, Terraform, Redis, Prometheus and OIDC.

🧠 Currently focused on governed AI runtime boundaries: MCP tool governance, intent resolution, OIDC workload identity, Redis-backed quotas, Prometheus-driven policy inputs, cost governance, risk scoring and audit.

"AI infrastructure should be observable, governable and boring in production."

Cloud Native Rockstar 2026

• 🧱 Build cloud-native and AI-native platforms with Kubernetes, GitOps and Infrastructure as Code
• 🧠 Design AI runtime and control-plane layers for private LLM inference, MCP tool calls, intent routing, fallback and autoscaling
• 📡 Implement OpenTelemetry-based observability for infrastructure and GenAI workloads
• 🛡️ Build governance workflows for identity, policy packs, prompt security, cost control, risk scoring, approvals, audit and sovereign AI
• 🎯 Architect GitOps delivery with Argo CD, Argo Rollouts, Helm and Terraform

⚠️ Fun fact: the best infrastructure is still the one nobody notices during business hours.

Two repositories demonstrate a complete enterprise reference architecture for governed private AI workloads:

flowchart TB
Users["Users / OpenAI SDKs / Agents"] --> Gateway["Execution Plane\nOpenAI Gateway"]
Agents["Agentic workloads"] --> Intent["Intent Proxy\n/v1/intent/resolve"]
Gateway --> Intent
Gateway --> MCP["MCP Gateway\nGoverned tool calls"]
Gateway --> Models["Model Backends\nOllama · vLLM · KServe"]
subgraph Control["Control Plane"]
Policy["Policy Packs"]
Identity["OIDC / JWKS Identity"]
Quota["Redis Tenant Quotas"]
Cost["Cost Governance"]
Risk["Risk Scoring"]
Approval["Human Approval Gate"]
Audit["Audit + Response Evaluation"]
end
Intent --> Policy
MCP --> Policy
Gateway --> Policy
Policy --> Identity
Policy --> Quota
Policy --> Cost
Cost --> Risk
Risk --> Approval
Approval --> Audit
Prom["Prometheus\nlive SLO + telemetry inputs"] --> Policy
Redis["Redis\nshared quota state"] --> Quota
Keycloak["Keycloak\nworkload identity"] --> Identity
Audit --> Obs["Observability\nGrafana · Loki · OpenTelemetry"]

AI Infrastructure OS control plane for governed private AI.

• Governance pipeline: policy pack → prompt security → quota → registry → cost → risk → approval
• Intent engine: natural-language request → agent/model/tools/region execution plan
• MCP tool registry, agent registry and signed model registry
• Redis-backed tenant quota and Prometheus live governance inputs
• Keycloak OIDC / JWKS identity, audit trail, response evaluations and sovereign AI checks
• Enterprise demo: Control Plane + Execution Plane + Ollama + Redis + Prometheus + Keycloak

AI Infrastructure OS execution plane for inference, tools and governed runtime traffic.

• OpenAI-compatible gateway with health-aware, cost-aware, fallback and canary routing
• Governance enforcement through CONTROL_PLANE_URL
• MCP gateway for governed tool calls
• Intent resolve proxy for agentic workflows
• OIDC/JWKS verification and workload identity forwarding
• Redis-backed tenant attribution, Prometheus metrics, vLLM, KServe, KEDA and GitOps

Together, they show a complete AI Infrastructure OS: the Execution Plane runs inference and tool calls, while the Control Plane governs identity, policy, cost, telemetry, audit, agents and intent.

Earlier hands-on work in cloud automation, GitOps, security and platform reliability:

• 🔄 Self-Healing Infrastructure with Chaos Engineering
  Kubernetes + LitmusChaos + Prometheus — auto-recovery pipelines and dashboards.

• 📦 GitOps Duel: ArgoCD vs Flux
  Side-by-side GitOps deployment comparison with ArgoCD and FluxCD on Kind.

• ☁️ Multi-Cloud IaC with Terraform + Terragrunt
  Reusable infrastructure stacks across AWS and Azure using Terragrunt modules.

• 🔍 AWS Security Audit with Prowler
  Automated scanning with Prowler + integration with Security Hub + GitHub Actions.

• 📊 Cloud-Native GitOps Platform with ArgoCD, Terraform, Monitoring & Security
  Prometheus, Loki, Grafana and Jaeger setup with alerting and dashboards.

💡 Want more? Visit github.com/justrunme?tab=repositories for future experiments.

🔭 Open to collaboration on:

• Platform Engineering / Developer Experience
• AI Infrastructure Architecture
• Private LLM Runtime Platforms
• GenAI Observability and Runtime Governance
• Kubernetes Operators / Controllers
• Cloud-native compliance & security
• Multi-cloud architecture (AWS / Azure / GCP)

🌍 Visit my Lab → Self-Healing Infrastructure with Chaos Engineering
for tools, experiments, and ideas that shouldn't run as root.

• 🧠 AI Infrastructure OS with Control Plane + Execution Plane architecture
• 🧩 MCP and Intent Governance for agentic tool calls and execution plans
• 🔐 OIDC/JWKS Workload Identity for governed private AI platforms
• 📊 Redis + Prometheus Governance Inputs for live quota and SLO-aware decisions
• 🧠 AI Runtime Decision Engines for model routing, fallback, health and cost-aware inference
• 📡 OpenTelemetry GenAI Observability for traces, metrics and runtime-level AI signals
• 🧭 AI Infrastructure Control Planes for governance, forecasting, approvals, audit, intent and policy updates
• 🛡️ Policy-Driven AI Governance with OPA, Rego, Conftest and GitOps workflows
• 🛡️ eBPF for observability and zero-trust runtime security

• 🤖 AI Infrastructure OS, inference routing, MCP gateways, intent engines, KServe, vLLM and KEDA
• 📡 OpenTelemetry, GenAI observability, Grafana and Loki
• 🧭 AI governance, identity, policy packs, cost governance, risk scoring, audit and approval workflows
• 🔄 GitOps, Helm, Argo CD, Argo Rollouts and Terraform
• ⚙️ CI/CD with GitHub Actions and GitLab CI
• 🛡️ Secure CloudOps and SRE practices
• 📬 Chat with me on Telegram → @justrunme