Dmytro M.

I don't just design systems, I run them: as a Senior DevOps Engineer with 5+ years in DevOps Engineering, I operate container platforms in production, the cloud and private networks underneath them, 27 ARM industrial devices in the field, and phone lines where an AI Automation agent answers real callers. Running them is where you meet the failures nobody plans for - servers you're afraid to reboot, a VPN that runs fine and then degrades until someone restarts it. Infrastructure rarely fails because the technology was wrong. It fails because nobody kept operating it after it was built. ▸ CONTAINERS, DOWN TO THE INTERNALS As a DevOps Engineer, I run Docker, Compose, and production Swarm - plus Git, GitLab CI/CD, and Jenkins pipelines where I accelerated BuildKit builds 6.2x (from 48m to 7m 40s) for Node.js and React workloads. Especially Containerization and networking across hosts - overlay networks, cross-host service discovery, routing between servers, and PostgreSQL Patroni HA clusters with sub-11s failover. That is where most Docker work stops, and where the expensive bugs live: services that resolve each other on one node and silently fail on the next. ▸ CLOUD AND ORCHESTRATION Working as an AWS DevOps Engineer on DevOps Kubernetes and Amazon Web Services (plus Google Cloud Platform and Amazon ECS), I handle Deployment Automation with Terraform, ArgoCD, Helm, and GitOps delivery. The interesting part isn't getting a deployment to start - it's probes that fail honestly instead of flapping (eliminating 100% of rollout 502/504 errors at 50 req/s), cutting AWS spend by 52% ($9.6k/mo) with Karpenter Spot and Graviton, storage that survives a node going away, and a rollback that has actually been tested. ▸ NETWORKS AND SECURE ACCESS As a Linux DevOps Engineer, I manage private networks, WireGuard and OpenVPN, firewalls, routing, DNS and TLS, access for your team and your CI, and Configuration Management with Ansible and HashiCorp Vault (rotating 1,200+ dynamic DB leases daily with SPIFFE mTLS). Including the diagnoses nobody enjoys - MTU and fragmentation, asymmetric routes, NAT: where everything looks configured correctly and the link still degrades over hours. In observability, I cut Prometheus TSDB cardinality by 92% (31 GB to 4.8 GB RAM) and reduced incident MTTR from 52m to 8m. ▸ AI, NOT JUST PLUMBING The whole range of AI Automation, from raw model APIs up to agent harnesses: context engineering, tool design, and memory - the one that decides whether anyone keeps using an automation. I run OpenAI, Anthropic, Google and local models behind one Python gateway with sub-140ms failover and 41% lower token spend via Redis semantic caching, so a provider change isn't a rewrite. ▸ AUTOMATION THAT HOLDS UP As a DevOps Engineer building AI Automation, I use n8n and Make, CRM and API integration, private AWS Bedrock RAG over your own documents (blocking 100% of PII leaks), and Python or JavaScript where no-code stops. The hard part is three in the morning: an API rate-limits you, a webhook fires twice. I build for that - retries, idempotency, alerts when a run fails. ▸ VOIP AND REAL-TIME VOICE Asterisk at the SIP, RTP and AMI level rather than through a web panel - trunks, call flows, one-way audio, registration failures, NAT. I built and run a real-time AI voice agent on Asterisk handling live calls across 5-7 concurrent SIP lines with sub-700ms latency: interruption handling, tool calls mid-conversation, transfer to a human. ▸ TOOLS I PUBLISH wireguard-tray, because WireGuard ships desktop clients for every platform except Linux. mnemo, persistent memory for agents that forget what you told them last week. dm-aioaiagent, the agent library I reuse across my own projects. All public on GitHub, so you can check them yourself before hiring a DevOps Engineer. ▸ EDUCATION MSc (Master's) in Computer Science. ▸ HOW I WORK As a DevOps Engineer, I back up before I change anything and verify the rollback path before I start. What I build runs on your servers, under your accounts, documented as I go. You own it - nothing is locked to me. I work best with teams that have a real system in production and want it understood and stable, not patched again. Tell me what's broken, or what you need standing up. I'll tell you honestly whether I'm the right person for it.