// platform · devops · observability · ai infra

Yaşar Orçun Kurnaz
Building reliable, observable, automatable infrastructure.

DevOps / Platform engineer with 20+ years across telecom, finance and education — Kubernetes & OpenShift, enterprise observability & AIOps, GitOps/Infra-as-Code, and local-first AI infrastructure.

Location Istanbul, TR Email tarcin@tarcins.network GitHub yokurnaz Labs api-rustfs.tarcins.network

Cloud & Containers

Kubernetes, OpenShift, Ceph/Rook/Longhorn storage, HA + DR design.

Observability & AIOps

CA Spectrum/UIM, BMC AIOps, Grafana, Zabbix — proactive detection & remediation.

Automation & GitOps

Ansible, Terraform, n8n, GitHub Actions, BlueBuild, immutable OS images.

Local-First AI Infra

LLM gateways, LM Studio, prompt-cache-aware compression, self-hosted model routing.

// selected work

Projects & Systems

🛰️

tarcinsblue — personal Fedora Atomic image

Immutable, signed Fedora Silverblue OS image built with BlueBuild. GitOps: every change lands as a PR into a GitHub Actions pipeline emitting a Sigstore/cosign-signed image, rebased via rpm-ostree.

BlueBuildGitHub ActionscosignFedora Atomic
☸️

Multi-node Kubernetes lab on IPG

Self-hosted multi-node cluster (kind-based) with RustFS distributed storage for persistent volumes, exposed via dedicated console & API endpoints — a real running control-plane.

KubernetesRustFSstorageself-hosted
🚀

OmniRoute AI gateway — local + free-tier LLM routing

Self-hosted LLM gateway routing one endpoint across local LM Studio models and 90+ free cloud tiers, with stacked RTK + Headroom lossless compression that preserves the prompt cache.

OmniRouteLM StudioRTK+Headroomprompt cachefree-tier routing
🏍️

KTM 1290 SAS dashboard — vehicle telemetry

Automotive reverse-engineering: read the KTM CAN bus (ECU 0x7EA) and decode TPMS at 433 MHz. Kotlin/Android dashboard fed by CAN-to-BT/USB bridges and an ESP32+MPC2515 interface.

CAN busKotlinESP32+MPC2515AndroidTPMS
🖥️

Self-hosted Hermes agent stack

Containerized personal agent (WebUI gateway + dashboard + terminal) on IPG, fronted by oauth2-proxy, wired to an LLM gateway via MCP — scheduled jobs and a real terminal + browser.

Dockeroauth2-proxyMCPself-hostedagent
🌡️

ESP32 telemetry & smart fan controller

ESP32-S3 fan controller with a web UI — 25 kHz PWM on GPIO7, WS2812 status LED, WiFi, mDNS/CDCOnBoot. Plus a 4S3P Li-ion battery-pack engineering plan for an EV bench.

ESP32-S3PWMWiFi/BLEweb UIpower
🤖

Self-hosted automation & agent layer (n8n + MCP)

Automation and agent platform hosted at n8n.tarcins.network: live status API endpoints for the Kubernetes & PostgreSQL clusters, an SSH command runner across the lab, an MCP server exposing web search, and an LLM agent that chats with the Postgres store.

n8nMCPwebhooksLLM agentautomation
n8n automation editor
📊

Central observability — Grafana over the self-hosted stack

Central Grafana dashboard aggregating health and metrics from the Kubernetes cluster and a cloud-native, high-availability PostgreSQL (operator-managed) database — fed by the lab's live status endpoints for real-time, single-pane monitoring.

GrafanaPrometheusHA PostgreSQLKubernetesobservability
Grafana monitoring dashboard
// toolkit

Stack

PLATFORMS
  • Kubernetes / OpenShift / OKD
  • Docker, Proxmox, KVM, VMware
  • Linux (Fedora Atomic, RHEL)
  • Cloudflare / self-hosted tailnet
  • Self-hosted secrets & credential management
OBSERVABILITY & AIOPS
  • Grafana / Prometheus
  • CA Spectrum, UIM, PM, NetQoS
  • BMC AIOps, Discovery
  • Zabbix, n8n, Flowise
AUTOMATION & AI
  • Ansible, Terraform, GitOps
  • GitHub Actions, BlueBuild
  • Python, Bash, YAML
  • LM Studio, OmniRoute, Ollama

Let's talk infrastructure

Open to platform / DevOps / observability engineering roles and principal consulting engagements.