Hi, I'm Aslan Tishbek
DevOps & AI Infrastructure Engineer
I build and automate cloud-native infrastructure using Kubernetes, GitOps, and IaC. I run a fully declarative bare-metal homelab with local AI workloads — proving production-grade skills at home.
About Me
I'm a self-driven infrastructure engineer passionate about building reliable, automated, and observable systems. My homelab is my proving ground — it runs the same tools used at scale in production: FluxCD, Terraform, Prometheus, and external-secrets, all on bare metal.
Beyond infrastructure, I'm deeply interested in on-premise AI — running LLMs and diffusion models locally with full control over the stack. I believe the best engineers understand the entire system from silicon to service.
Certified CKAD and HashiCorp Terraform Associate (004). Currently targeting AWS Solutions Architect as the next milestone.
Homelab Specs
- CPURyzen 7 5800X3D
- GPURX 7900 XTX 24GB
- RAM48GB DDR4
- Storage2× 2TB NVMe
Quick Info
Technical Skills
Container & Orchestration
GitOps & IaC
Cloud & Networking
Observability
AI / Local Inference
Languages & Tools
// full stack
Production-Grade Homelab
A fully declarative, GitOps-driven infrastructure running on bare metal — mirroring real-world SRE/DevOps practices.
Homelab Architecture
Hosted at a-tishbek.info
Infrastructure as Code
-
Ansible Configuration Management
Automated baseline OS setup, user groups, K3s bootstrapping, and GPU drivers across the entire homelab.
-
Terraform Provisioning
AWS S3 state backend. Provisions namespaces, Route53 DNS wildcard records, and base Helm charts via modules.
-
FluxCD GitOps
Continuous reconciliation of the Kubernetes state directly from GitHub. Zero manual
kubectl apply. -
External Secrets Operator
Integrated with Bitwarden to inject API keys and credentials securely into the cluster without Git exposure.
Bare-Metal AI Stack
// AI workloads run as systemd services to bypass container overhead for massive models
-
Ollama — LLM Engine
Qwen3 family (8B / 32B / 30B-Coder) on ROCm, fronted by LiteLLM and served to Open WebUI and a fleet of Claude Code subagents.
-
ComfyUI — Diffusion & Video
Running Stable Diffusion 3.5 Large and WanVideo generation natively on the 24GB AMD GPU.
-
whisper.cpp — Speech-to-Text
Native HIP build of
large-v3on the RX 7900 XTX, exposed as an OpenAI-compatible/v1/audio/transcriptionsendpoint atstt.a-tishbek.info. ~7.6× realtime on the 7900 XTX. -
Qdrant + local-rag MCP
Vector store deployed via GitOps with a custom MCP server using
nomic-embed-textfor semantic codebase search across homelab repos. -
doc-cache MCP + Smart-Read Hook
SQLite-backed cache fronting Context7 library docs and Ollama answers, plus a PreToolUse hook that intercepts oversized file reads and routes them to local summarizers.
-
Ollama-Backed Agent Fleet
~30 specialist subagents (code review, log triage, vision OCR, k8s/GitHub/Gmail/Drive payload condensers) routing bulk work off Claude Opus onto local Ollama via custom MCP servers.
// public openai-compatible endpoints
litellm.a-tishbek.info/v1
— chat completions (Qwen3 fleet via LiteLLM router)
stt.a-tishbek.info/v1
— audio transcriptions (whisper.cpp large-v3 on ROCm)
Observability & Networking
-
Traefik Ingress & TLS
Wildcard routing for
*.a-tishbek.infosecured with Let's Encrypt via cert-manager HTTP-01 challenges. -
kube-prometheus-stack
Prometheus, Alertmanager, and Grafana providing full node and pod-level metrics with custom dashboards.
-
Loki + Promtail
Centralized log aggregation across all cluster workloads, queryable from Grafana alongside metrics.
-
AdGuard Home — Network-Wide
Running on hostNetwork to intercept all LAN DNS and serve DHCP, bypassing ISP router limitations.
Services Running
Goals & Roadmap
Certifications
Engineering Roadmap
Updated May 2026// recently shipped
large-v3 on ROCm; OpenAI-compatible /v1/audio/transcriptions at stt.a-tishbek.info.// up next
Get In Touch
I'm open to DevOps, SRE, and infrastructure engineering opportunities. Whether it's a full-time role, freelance project, or just a tech conversation — reach out.