Solution architecture
Designing systems before building them: trade-off decisions, disaster-recovery planning, migrations, design reviews and written standards, and mentoring engineers along the way.
Den Holtsev
DevOps & Infrastructure Engineer
Building infrastructure that stays up.
I architect production infrastructure and see it through to code, rollout and on-call: Kubernetes on bare metal and Proxmox VE, DNS and web edge, AWS and the pipelines behind them, now with AI agents in the loop.
- 8+ yrsin production infrastructure
- Hybridbare metal, Proxmox and AWS
- IaCevery change versioned and reviewed
- AIagents in daily ops, with guardrails
01 · Expertise
What I'm good at
Hands-on ownership in production, from architecture to day-to-day operations and on-call.
On-premise platforms
Kubernetes on bare metal and Proxmox VE: cluster builds and upgrades from versioned installer images, node changes through pipelines, load balancing, persistent storage, and VM provisioning from code.
DNS & web edge
DNS and web services that stay fast under heavy load and attacks: authoritative and recursive DNS, reverse proxies, load balancing and TLS termination, with filtering, rate limiting and high availability.
AWS & Infrastructure as Code
AWS infrastructure in Terraform (networking, load balancing, containers, managed databases, IAM), plus fleet configuration with Chef/Cinc and Puppet: reusable roles, tested changes, gradual rollouts.
Delivery & observability
One shared pipeline flow for infrastructure code (plan on pull request, apply on merge, short-lived credentials, secret scanning), plus metrics, dashboards and alerts managed as code.
AI-assisted operations
AI agents connected to metrics, logs, tickets and pipelines; repositories prepared for agents; AI review in CI. All inside guardrails, so people stay in charge of production.
02 · AI in practice
How I use AI in infrastructure work
Practical, not hype: agents work with real metrics, logs, tickets and pipelines, and every change still goes through review.
Agents on real tooling
AI coding agents connected through MCP to monitoring, logs and ticketing, so incident triage starts from real data instead of guesses.
Runbooks as agent skills
Recurring procedures turned into reusable agent skills, so routine checks and investigations follow the same documented steps every time.
Agent-ready repositories
Repositories get agent instructions with their conventions and ordered checks (lint → test → build), pre-approved safe commands and a deny-list for secrets.
AI review in CI
An AI reviewer checks every pull request alongside human reviewers and catches routine issues before a person has to.
Guardrails
- Read-only access by default
- Secrets kept out of agent context
- Production changes go through review and approval
03 · Stack
Tools I use in production
Mostly open source, picked for the job rather than the hype.
Kubernetes & virtualization
DNS, web & networking
AWS
IaC & config management
CI/CD & packaging
Observability
AI tooling
Data, messaging & secrets
Languages
04 · Contact
Have a project or role in mind?
Drop me a line and tell me a bit about it. Email is the fastest way to reach me.