I build reliable platforms where Kubernetes, CI/CD and private LLMs work as one production system.

 👋 About    🧭 How I work    🏗 What I build    🧰 Toolbox    ✉️ Contact 

_{Click a section to jump — and open the ▸ arrows below, they hide a bit more.}

About

I came to AI from systems administration, production operations and DevOps leadership. That shapes how I build: reproducibility, isolation, observability, security, cost and recovery come before demo magic.

My focus is the engineering layer between infrastructure and AI — self-hosted inference, RAG, agent runtimes, MCP integrations, code intelligence and delivery automation. I turn these pieces into systems that a team can deploy, operate and trust.

How I work

Every engagement follows the same loop — from a fuzzy business problem to something that runs, is observed and keeps getting better:

flowchart LR
    A["🎯 Business<br/>problem"] --> B["📐 Architecture<br/>and trade-offs"]
    B --> C["⚙️ Platform<br/>Kubernetes · CI/CD"]
    C --> D["🧠 AI layer<br/>inference · RAG · agents"]
    D --> E["🔭 Operate<br/>observe · audit · recover"]
    E -. "feedback" .-> A

Loading

Six principles decide almost every design choice:

🔁 Reproducibility If it can't be rebuilt from code, it isn't production.	🧱 Isolation Every agent, task and tenant gets its own boundary.	🔭 Observability You can't operate — or improve — what you can't see.
🔐 Security SSO, RBAC, secrets and audit are day-one features.	💸 Cost GPU hours and tokens are budgeted like any other resource.	🛟 Recovery A backup only counts once a restore has been rehearsed.

What I build

🏗  Production platforms — environments that stay up and can be rebuilt

• Kubernetes and Docker environments
• GitLab CI/CD pipelines and traffic routing
• Monitoring, audit, backup and recovery

🧠  Private AI infrastructure — models that run on your hardware, under your rules

• Local model serving with OpenAI-compatible endpoints
• GPU capacity planning and long-context workloads
• Air-gapped deployment

🤖  Agentic engineering systems — agents with sandboxes, not superpowers

• Isolated execution sandboxes
• Tool access and MCP servers
• Repository intelligence and AI-assisted operations

🛡  Enterprise foundations — the boring parts that make it adoptable

• SSO / RBAC and secrets management
• Traceability and safety controls
• Operable upgrade paths

🧭  Technical leadership — from ambiguity to a running service

• Architecture and roadmaps
• Hands-on delivery, from an ambiguous business problem to a running production service

Toolbox

Same toolbox, as a plain list

Platform & delivery	Kubernetes · Docker · GitLab CI/CD · Linux · NGINX/HAProxy · PostgreSQL · Redis
AI engineering	Ollama · vLLM · OpenAI-compatible APIs · MCP · RAG · Qdrant/FAISS · agent orchestration
Security & operations	Keycloak/SSO · RBAC · secrets management · audit logging · ELK/Logstash · Zabbix · backup/restore
Software engineering	Go · Python · Bash · TypeScript · React · FastAPI

Current focus

• 🧱 Private agent platforms with isolated per-task sandboxes and Docker/Kubernetes delivery.
• ⚡ Reliable local LLM inference for coding, operations and long-running agent workflows.
• 🔍 AI-assisted CI/CD governance, infrastructure diagnostics, code intelligence and technical documentation.

🇷🇺  Коротко по-русски

Я пришёл в AI из системного администрирования, эксплуатации production-систем и DevOps-лидерства. Поэтому воспроизводимость, изоляция, наблюдаемость, безопасность, стоимость и восстановление для меня важнее демо-магии.

Занимаюсь инженерным слоем между инфраструктурой и AI: self-hosted-инференс, RAG, агентные рантаймы, MCP-интеграции, анализ кода и автоматизация поставки. Собираю из этого системы, которые команда может развернуть, эксплуатировать и которым может доверять.

Что строю: production-платформы на Kubernetes и Docker · приватную AI-инфраструктуру · агентные системы с изолированными песочницами · enterprise-основу (SSO/RBAC, секреты, аудит) · и руковожу технически — от размытой бизнес-задачи до работающего сервиса.

Contact

Need infrastructure that survives production — and AI that can safely operate inside it?

I am open to architecture reviews, technical leadership and hands-on collaboration on DevOps, private AI and agent-platform projects.

_{Hand-written SVG animations and native GitHub Markdown — no widgets, no trackers.}