DevOps & Cloud Architect with 10+ years of experience building scalable infrastructure on AWS, Azure, GCP, and other cloud platforms.
Expertise: Infrastructure as Code (Terraform, Terragrunt, Helm), Kubernetes, CI/CD pipelines, monitoring & observability, and cloud architecture design.
AI-augmented engineering: coding agents (Claude Code, Cursor, Copilot), MCP integrations, and agentic workflow development applied to DevOps automation.
Track Record: Led technical teams, designed cloud architectures from scratch, launched 10+ production systems end-to-end, and mentored engineers across multiple organizations.
AWS Community Builder and certified cloud professional passionate about automation, DevOps best practices, and sharing knowledge through talks and technical content.
Tech Lead / Senior DevOps Engineer • Naviteq Ltd • Jan 2022 – Present
- Delivered 10+ client projects end-to-end, from architecture design to production.
- Architected and deployed highly available, secure AWS infrastructure with cost optimization (saved up to 40%).
- Standardized reusable Terraform/Terragrunt templates and CI/CD pipelines across teams.
- Operated Kubernetes clusters and authored Helm charts to streamline deployments.
- Built monitoring and logging stacks with Prometheus, Grafana, Loki, and OpenSearch.
- Mentored engineers through code reviews and architecture guidance.
Senior DevOps Engineer • OneSoil • Aug 2021 – May 2022 (consulting)
- Designed a multi-account AWS architecture aligned to security and scalability best practices.
- Reduced cloud costs by ~30% with no loss of performance.
- Built a Kubernetes platform for scalable deployments.
- Added observability with Prometheus and Grafana.
Lead System Engineer • EPAM Systems • Apr 2016 – Aug 2021
- Grew from junior to lead engineer over five years.
- Designed AWS infrastructure and automated it with Terraform and Ansible.
- Built and maintained CI/CD pipelines with GitLab CI and Jenkins.
- Led the team as tech lead: planned work, assigned tasks, and reviewed solutions.
- Migrated enterprise systems to the cloud (AWS, Azure).
- Delivered several international projects across fintech and retail.
Senior Network Engineer • RUE Beltelecom • Mar 2011 – Apr 2016
- Supported clients consuming Internet, VoIP, and IPTV services.
- Maintained the backbone network and corporate clients (VPN, Ethernet, SHDSL).
- Automated network equipment maintenance with custom scripts.
- Cloud Architecture: AWS (security, networking, multi-account, FinOps), Azure, GCP
- Platforms: Kubernetes, ArgoCD, Helm; GitOps at scale
- AI Engineering: AI-assisted development (Claude Code, Cursor, Copilot), MCP servers & integrations, agentic workflows & LLM-driven automation
- CI/CD: GitHub Actions, GitLab CI, Jenkins; trunk-based workflows
- IaC: Terraform, Terragrunt, CDK, Pulumi; reusable modules, policy-as-code
- Observability: Prometheus, Grafana, OpenTelemetry; actionable SLOs and alerting
- Security: IAM hardening, secret management, SBOM, supply chain controls
- Cost: usage analytics, right-sizing, autoscaling, waste elimination
- Ways of Working: pragmatic leadership, enablement, knowledge transfer
- Engineer, Belarusian State University of Informatics and Radioelectronics (2015)
- Information technologies and management in technical systems
- Ongoing professional education through certifications and community work
- The node that boots, bills, and never joins Sep 9, 2026
Karpenter picks the instance type from one file and the AMI from another, and nothing checks that the two agree. EC2 now lets an AMI owner declare which instance types it supports, turning fifteen minutes of silence into a rejected launch.
- Rootless Kubernetes went on by default, and nothing happened Sep 9, 2026
The KubeletInUserNamespace gate reached beta in Kubernetes 1.37 and now defaults to on, yet nothing changes on an existing cluster. What the gate really does, how it differs from hostUsers: false, and where rootless nodes are worth running today.
- Karpenter 1.14 makes headroom an object, and the docs are wrong about it Sep 2, 2026
Karpenter 1.14 replaces the pause-pod headroom hack with a CapacityBuffer resource, but the concept page documents an API version the CRD does not serve and disagrees with it on how the buffer is sized.
- Terragrunt read one of your two dependency blocks and said nothing Sep 2, 2026
Two dependency blocks sharing a label resolved silently to the last one for years; Terragrunt 1.1.4 adds a strict control that turns the shadowing into an error you can enforce in CI.
- HPA scale-to-zero shipped. The CPU limit is the interesting part Aug 27, 2026
Kubernetes 1.37 lets an HPA hold a workload at zero pods, but only on object and external metrics, and the reason why decides which of your workloads it is actually for.
- Terraform 1.16 let import into modules. Read the review thread before you celebrate Aug 27, 2026
Terraform 1.16 allows import blocks inside modules, which fixes brownfield adoption and quietly moves review burden into every module version bump you approve.
- Your new IAM role starts at PowerUser Aug 18, 2026
IAM role manager creates the role for you as you build. Four of the six supported consoles pull the same PowerUserRoleTemplate, so the starting point is close to full service access. What it grants, the SCPs that stop it, and how to find the roles it already made.
- The plan wanted to replace the production database Aug 12, 2026
Swapping an RDS module changed the resource address in state, so Terraform planned a destroy on a live database, and one state mv was the entire fix.
- CloudWatch came for your collector Aug 4, 2026
CloudWatch now scrapes Prometheus metrics from EKS, ECS, EC2, MSK and OpenSearch with no collector of yours in the path. The operational win is real. The cost just moved onto three separate meters.
- destroy = false is not prevent_destroy Aug 4, 2026
Terraform 1.16 beta adds lifecycle destroy = false. It reads like prevent_destroy and behaves like the opposite: the resource leaves state and keeps running, and the plan output never calls that a leak.
- Terragrunt 1.1: six experiments became defaults Aug 4, 2026
Terragrunt 1.1 turned six --experiment flags into defaults, including stack dependencies and the content addressable store. Here is what autoinclude changes, and what moved under your CI without a pull request.
- AI on-call is confident. That's exactly the problem. Jul 23, 2026
AI advice made people less accurate but more confident, collapsing their willingness to say 'I don't know' from 44% to 3%. On-call, that eroded doubt is the danger, and read-only, approval-gated assistants are how you keep it.
- EKS can roll back an upgrade now. Read the fine print before you trust it. Jul 23, 2026
EKS can now revert a control-plane upgrade from N to N-1 within 7 days, but the kubelet skew rule means only a phased upgrade keeps that rollback within reach.
- The half of your EKS bill nobody reads Jul 3, 2026
Everyone optimizing their EKS bill tunes compute and ignores the other half of the invoice: data transfer, hiding in a dozen tiny cross-AZ and NAT rows.
- One field in Kubernetes 1.34 vs your cross-AZ bill Jul 3, 2026
Kubernetes 1.34's traffic-distribution change reads as a networking refinement. It's also a one-field lever on the AWS cross-AZ data-transfer charge you're probably still paying.
- AI Code Security Is a Runtime Problem, Not a Review Problem Jun 26, 2026
30% of developers knowingly ship vulnerable AI code to production. The fix isn't better review, it's a three-layer trust boundary that treats runtime as the real safety line.
- Kubernetes probes are a mental model, not a checklist Jun 21, 2026
What startup, readiness, and liveness each actually control, and why treating probes as three boxes to tick is how people take their own services down.
![[Field Kitchen] DevOps Kitchen Talks on Zed Conference (Eng)](https://i1.ytimg.com/vi/Ptwk-rKQ5WM/mqdefault.jpg "[Field Kitchen] DevOps Kitchen Talks on Zed Conference (Eng)")
![30 глупых вопросов DevOps-инженеру [Карьера в IT]](https://i1.ytimg.com/vi/vdHb6U3AH9E/mqdefault.jpg "30 глупых вопросов DevOps-инженеру [Карьера в IT]")
