Senior DevOps Engineer · Kubernetes, GitOps & reliability
I build and operate delivery platforms for fintech systems, and contribute fixes to the tools behind them. Six years in production infrastructure; currently at Innotech, based in Astana, Kazakhstan.
LinkedIn · Email · Contribution notes
These changes were merged upstream, and the release is noted where one already ships them. Each link opens the implementation, tests, and maintainer discussion.
| Project | What I contributed |
|---|---|
| Helm · #32281 | Implemented Go parsing of GnuPG pubring.kbx keyrings for chart provenance verification, with bounds checks, malformed-input tests, and legacy-keyring compatibility. Released in Helm v4.3.0. |
| OpenTelemetry Collector · #50507 | Finished removing the Elasticsearch exporter's deprecated flush and num_workers options, tracked since 2025 in #42718. Old configs are now rejected instead of being migrated automatically; added tests and a breaking changelog entry. Released in v0.162.0. |
| KubeAI · #733 | Fixed the Ollama startup probe, which could pass without pulling or preloading the model (#721). Rewrote the probe script, added regression tests that execute the generated probe, and extended the PVC end-to-end test. Released in v0.23.5. |
| Flux source-controller · #2127 | Added OCI source revisions to event metadata so notification consumers can correlate artifacts with source commits; included controller tests and documentation. |
| Flux CLI · #5885 | Added --add-ignore-paths to artifact build, push, and diff commands to extend existing exclusions; covered flag combinations with tests. |
Also contributed test cleanup to the OpenTelemetry Collector. Browse my merged pull requests →
- Migration: led the move of 100+ microservices from OpenShift to Kubernetes across 10 clusters serving 15,000+ RPS, with a zero-downtime production traffic cutover.
- Delivery: replaced Jinja2/Ansible deployment flows with versioned Helm charts and unified rollback procedures; built a pipeline that halved publication time.
- Efficiency: reduced private-cloud resource consumption by 25% while maintaining a 99.9% SLO.
- Operations: Istio mTLS and authorization policies, TLS rotation, production on-call, blameless incident reviews, and mentoring seven engineers.
helm-hook-preflight: an alpha Python CLI that reads a rendered Helm bundle and flags pre-install hook Jobs that depend on a ServiceAccount or Secret the chart creates too late for a fresh first install. Works offline, without cluster access. Listed in Kubetools under testing tools.
Crewboss: an experimental framework for coordinating coding agents with role-based permissions, task orchestration, a web dashboard, and review/CI gates. The repository contains the reference runtime, tests, and architecture notes.
Contribution Notebook: a local Python CLI for collecting public GitHub activity, writing Markdown reports, and keeping contribution plans alongside evidence. Includes an offline demo, tests, and CI; GitHub access is read-only.
Quarter: a personal finance app in development. Building the native iOS client with SwiftUI and SwiftData, a transactional outbox for sync retries, and budget indicators that compare spending with elapsed time. Source is currently private.
Ops reproductions contains runnable examples from my community troubleshooting: kube-bench authorization checks with process fixtures, and Data Prepper metrics checked against an official container. Each example documents its pinned versions, expected results, and verification limits. Investigation notes →
| Area | Tools |
|---|---|
| Platforms | Kubernetes, OpenShift, Docker, Istio, Linux |
| Delivery & infrastructure | Helm, Terraform, Ansible, Jenkins, TeamCity, AWS, OpenStack |
| Observability & data | Prometheus, Grafana, Elastic Stack, Kafka, PostgreSQL, Redis |
| Code | Go for upstream contributions; Python, Bash, Groovy, Java, SQL |
Interested in Senior DevOps, SRE, and infrastructure roles: remote from Kazakhstan through B2B/EOR, or employer-sponsored relocation. English C1; Russian native.