Русская версия

Mikhail Basov

Михаил Басов, also written Michael Basov

Senior DevOps engineer at Sberbank. I work across the Risk division, about 1,500 microservices, and help its teams with releases, monitoring and Kubernetes.

I build release platforms and look after Kubernetes, monitoring and AI tooling. Outside the day job I research why alerts stay silent, send fixes to open-source monitoring projects and judge student hackathons.

Email me LinkedIn GitHub

Mikhail Basov

What I do

Platform work across the division

Support teams across the Risk division bring me questions about releases, monitoring and Kubernetes, and I help them sort things out. I also run 70+ production services and 6 LLM-based agents on Kubernetes myself.

Release automation

I built and own the release orchestration platform behind about 470 releases a month. Coordinating a release used to take 2–4 hours. Now it takes under 10 minutes.

AI in engineering

I started the AI code review service that checks 1,200+ merge requests a day, with a median response of about 30 seconds. It received Sberbank's internal Engineering Achievement of the Year award for 2025. I also build with AI agents every day: they work from a written plan, and every change gets an independent review before it merges.

Observability and alerting

I study alerting rules that stay silent or fire on the wrong condition, and fix them where they ship. My fixes are merged into Ceph, Rook and Grafana Mimir.

Everyday tools: Kubernetes, Helm, Jenkins and Groovy shared libraries, Prometheus and PromQL, Grafana, OpenTelemetry, PostgreSQL, Kafka, Python, Claude Code and Codex.

How I got here

I grew up in a small town in the Urals and moved to Yekaterinburg on my own at 15 to study information security. I started working in IT in 2021 as a systems administrator, then spent two years deploying enterprise security systems for large companies: data leak prevention and privileged access management.

In 2024 I moved to DevOps at Sberbank, and in 2026 I became a senior engineer. Today I work across the whole Risk division. The security background still shapes how I build platforms. Alongside work, I'm studying for a BSc in Software Engineering at Ural Federal University.

Research

Silent or Wrong: Producer-Contract Defects in Shipped Prometheus Alerting Rules

Alerting rules ship with exporters, Helm charts and monitoring mixins to people who did not write them. Many quietly depend on facts about their metrics, such as units, types, labels and histogram buckets, that the rule never states. When those facts are wrong, the alert never fires or fires on the wrong condition. I call these unstated facts the rule's producer contract, and the bugs they cause producer-contract defects. The study examines how often this happens in widely used open-source rules, where it comes from and whether current tools catch it.

I chose the topic, designed the study and ran it with LLM agents under my supervision.

Preprint coming in November 2026

Open source and community

Merged contributions

  • Ceph
  • Rook
  • Grafana Mimir
  • Kubernetes documentation
  • vLLM Production Stack
  • Argo CD mixin
  • awesome-prometheus-alerts

Hackathon judge

Judged the final round of the CSC Back-to-School Hackathon 2026, which drew 115 student projects. Coming up: judging the CSC Winter Sparkathon 2026 and eCYBERMISSION 2027.

Now

What I'm working on in October 2026.

  • The silent alerts study. Finishing the manuscript. The preprint comes out in November 2026.
  • Hackathon judging. On the jury of the CSC Winter Sparkathon, 30 November to 6 December 2026.
  • Upstream fixes. More alerting fixes are in review in Thanos, kube-prometheus and kubernetes-mixin.

Talks

So far my talks have been internal, for engineering teams at Sberbank. I'm open to conference talks, meetups, podcasts and panels, in English or Russian. Topics I can speak on:

  • Why your alerts never fire. How shipped alerting rules break on facts about metrics they never state, and how to catch it.
  • AI code review at bank scale. Running an LLM reviewer on 1,200+ merge requests a day in a regulated environment.
  • From hours to minutes. Building release orchestration for 470 releases a month.
  • AI-native DevOps. Letting agents do real engineering work with plans, review gates and audit logs.

Bio for organizers

Use either version on event pages and in announcements, together with this headshot.

Short

Mikhail Basov is a senior DevOps engineer at Sberbank and a researcher of production alerting.

Full

Mikhail Basov is a senior DevOps engineer at Sberbank, working across its Risk division of about 1,500 microservices. He built the release platform behind 470 releases a month and started an AI code review service that handles 1,200+ merge requests a day. He researches alerting rules that stay silent, contributes fixes to open-source projects such as Ceph and Grafana Mimir, and judges student hackathons.

Contact

Get in touch about talks and podcasts, hackathon judging, or open-source work on monitoring and alerting.

michael.s.basov@gmail.com

I work in English and Russian. Also on LinkedIn, GitHub and ORCID.