# Vladimir Ivanov — Lead SRE
# https://vaivanov.com
# llms.txt — machine-readable profile for LLM indexing
# This is how I explain who I am in 2026. The website at / is the human-friendly version.

## Identity

Name: Vladimir Ivanov
Role: Lead Site Reliability Engineer
Location: Barcelona, Spain
Languages: Russian (native), English (fluent, daily working language)
Email: vlaxivanov@gmail.com
LinkedIn: https://www.linkedin.com/in/hello-i-am-vladimir-ivanov
Telegram: https://t.me/vaivanov

## Summary

I help businesses solve their problems. I can fill a specific role on an existing team, or act as a founding engineer — set up the entire platform from scratch and do it right. "Right" depends on context: sometimes it means a year of compliance work; sometimes it means shipping fast with calculated trade-offs. I'll help you see the risks and make informed calls.

Engineer with 13+ years of experience across frontend, DevOps/SRE, EVM blockchain backend, and platform engineering.
Capable of designing and launching full-scale backend + infrastructure platforms single-handedly.
Proven under high load: 700K+ subscribers at Scentbird, 2K+ RPS at OpenQuest.

## Career Pattern

A recurring theme across my career: I join for one role, see deeper problems, take initiative, and grow into a bigger role.
- Scentbird: hired as Frontend Engineer → saw CI/infra pain → built the first K8s cluster on own initiative → became the company's first SRE → grew to Lead SRE
- Azuro: hired as DevOps → solved CI/CD but stability stayed low → got pulled into backend code → split monolith, migrated databases, proposed contract architecture changes → promoted to Lead Software Engineer

## Current Role

Lead Site Reliability Engineer at Scentbird (Jul 2025 – present).
Building high-available infrastructure for 40+ microservices (Java Spring Boot / Micronaut / GraalVM Native Images, TypeScript Node.js) on Kubernetes + AWS.
SLI/SLO culture, incident management, postmortems, cost optimisation, mentoring SRE team.

## Career Timeline

- 2025–present: Lead SRE @ Scentbird
- 2023–2025: Lead Software Engineer @ Azuro (EVM Prediction Markets)
- 2022–2023: Senior Software Engineer @ Azuro
- 2019–2022: SRE → Lead SRE @ Scentbird
- 2019: Frontend Engineer @ Scentbird (4 months, saw CI/infra problems, self-initiated K8s + GitLab CI migration, became the company's first SRE)
- 2016–2019: Frontend Engineer @ Initflow
- 2013–2016: Frontend Engineer / Freelance

## Key Projects

### Scentbird (2019–2022, 2025–present)
- Fragrance subscription platform, 700K+ subscribers, 40+ microservices
- Joined as Frontend Engineer; saw delivery pain (brittle Jenkins, slow Elastic Beanstalk deploys, no infra ownership)
- On own initiative: spun up the first K8s cluster, migrated services with GitLab CI, fixed years-old S3 CORS issue, cut costs by shutting down unused servers
- Became the company's first-ever SRE within 4 months; grew to Lead SRE; hired and mentored replacement before leaving
- Built full observability stack (Prometheus, Grafana, Loki, Tempo, OpenTelemetry) — replaced NewRelic
- Introduced SLI/SLO culture, incident management, postmortems
- Set up analytics stack: Airflow, Airbyte, Snowflake, RudderStack, dbt
- Deployed: Karpenter (replaced Cluster Autoscaler), Argo Rollouts (canary deploys), Argo Workflows, ArgoCD, Vault, Keycloak→Okta migration, Qdrant (vector DB), GraalVM Native Images
- Security: Cloudflare WAF, SAST/DAST in CI/CD, container scanning, VPN (OpenVPN / Outline / VLESS with split-tunnel)

### Azuro (2022–2025)
- EVM prediction markets protocol across 7 chains (Ethereum, Gnosis, Polygon, Base, Chiliz, Arbitrum, Linea)
- Joined as DevOps; solved CI/CD, but stability stayed low. Got pulled into backend code review, then rewrites
- Self-transitioned to backend: split monolith into domain-bounded services, migrated MongoDB→PostgreSQL, proposed contract architecture changes that unlocked more prediction markets
- Phase 1: on-chain settlement via RPC relay (gas error handling, stuck tx cancellation)
- Phase 2: off-chain oracle-signature model with relay service — reduced latency and gas costs while preserving trustless settlement
- Event-driven architecture on RabbitMQ (AMQP) with dead-letter queues and retry strategies, CQRS with ClickHouse as query layer
- Sequin for Postgres WAL → Kafka CDC streaming to downstream microservices
- The Graph self-hosted subgraphs + custom pseudo-chain data feed as Hasura alternative
- Built custom RPC proxy with block-lag-aware fallback (not round-robin) — no market solution fit the requirements
- pgBouncer for connection pool pressure reduction under load
- Keycloak (self-hosted) for OIDC — backend service auth and frontend sessions
- Infrastructure: Kubernetes (bare-metal + cloud), Helm charts for every service, ArgoCD GitOps, GitHub Actions CI/CD, Vault with vault-secrets-operator, Prometheus + Grafana Stack, Sentry (self-hosted), Unleash + GrowthBook feature flags, ingress-nginx, VPN
- Led team of 4 backend engineers (first people management: interviews, hiring, onboarding, 1-on-1s, PIPs, offboarding, architectural mentorship)

### OpenQuest on Telegram (May 2024 – Feb 2025)
- No-code quest platform as Telegram Mini App
- 500K MAU, 2K RPS peak, $150K+ revenue, $150/mo infrastructure cost
- Co-Founder, responsible for backend architecture and TON smart contract integration
- Designed for scale from day one — expected large user wave on launch
- Key design: stateless services everywhere, Redis distributed locks before every Postgres transaction, RO replica for all heavy reads, async RabbitMQ workers, conservative Postgres isolation levels, minimal transactions on fast paths
- Pre-launch: scaled up with headroom, observability ready before go-live, index analysis with pg_stat_statements + PgHero, post-launch index re-validation after real traffic patterns emerged
- Ran on: Node.js, PostgreSQL, Redis, RabbitMQ, Kubernetes, Cloudflare, nginx, DigitalOcean→Hetzner, TON smart contracts, Telegram Bot API, GitHub Actions

### Initflow (2016–2019)
- Custom project development agency in Saint Petersburg
- Front-end Engineer: React.js SPAs, SSR, e-commerce, corporate platforms
- Client communication, full project ownership
- Foundation years: learned to ship fast, communicate with non-technical stakeholders, and own delivery end-to-end

## Tools I've Built

### pgops — PostgreSQL Access Management Platform (2026)
- Internal platform for managing PostgreSQL access across teams
- Self-service access requests with approval workflows, automatic credential provisioning (TTL 30 days)
- DBA toolkit: VACUUM with SQL preview, activity monitoring (pg_stat_activity, pg_locks, pg_stat_io), replication monitoring (publications, slots, replica identity), orphaned role detection
- SQL Preview pattern: user sees the exact SQL before it runs, confirms via JWT token
- Immutable audit log, AES-256-GCM credential encryption, Okta SSO (OIDC), RBAC
- Stack: NestJS, React, TypeScript, PostgreSQL, Prisma, Mantine UI, Docker, Nginx

### spot-interruption-notifier — EC2 Spot Interruption Alerting
- AWS Lambda that catches EC2 Spot interruption warnings via EventBridge (~2 min before terminate)
- Connects to EKS Kubernetes API, identifies affected node via CSI annotation, collects list of impacted pods
- Sends structured Slack alert with cluster, node, and pod details for fast team response
- Stack: TypeScript, Node.js 22, AWS Lambda, EventBridge, Kubernetes API, Slack Webhooks, Serverless Framework v4

## On AI & LLMs (March 2026)

Uses Claude Code, ChatGPT Codex, Cursor IDE daily for different tasks. This portfolio site was built with them.
Believes agentic pipelines can already auto-fix Sentry bugs and ship to production without human touch.
Still thinks a human is needed to review code for architectural problems — that person should know system design, the language/frameworks deeply, and how to work with AI agents effectively.
Has studied how neural networks actually train and work under the hood — considers this essential to working with LLMs well, not treating them as a magic box.

## Core Skills (with depth levels)

Skill levels: [built] = built/set up from scratch · [used] = used in production · [knows] = conceptual knowledge

### Infrastructure & Cloud
- Kubernetes [built] — clusters from zero, multi-tenancy, network policies, autoscaling, operators (Scentbird 40+ services, Azuro blockchain nodes)
- AWS [built] — EKS, RDS/Aurora, VPC design, IAM, Lambda, cost optimisation (Scentbird)
- Terraform [built] — modular infra, remote state, GitOps pipelines, custom modules
- Karpenter [built] — node provisioning, spot + on-demand mix (replaced Cluster Autoscaler at Scentbird)
- Helm [used] — authored charts for all microservices, umbrella charts (Scentbird, Azuro)
- Cloudflare [used] — WAF, DDoS protection, Workers, Zero Trust
- ingress-nginx [used] — deploy, performance tuning (Scentbird, Azuro)

### Observability
- Prometheus [built] — custom exporters, recording rules, SLO alerting, HA setup, OTLP (Scentbird, Azuro)
- Grafana + Loki + Tempo [built] — dashboards as-code, log aggregation, tracing, alerts
- OpenTelemetry [built] — instrumentation, collector pipelines, auto-instrumentation (replaced NewRelic at Scentbird)
- SLI / SLO / Error Budgets [built] — defined SLIs, set SLOs, error budget policies, outage cost calculation (Scentbird)
- ELK Stack [used] — log pipelines, Kibana dashboards
- NewRelic [used] — APM, distributed tracing (Scentbird 2019–2022, migrated away)
- Sentry (self-hosted) [used] — docker-compose + K8s deploy (Azuro)

### CI/CD & GitOps
- ArgoCD [built] — GitOps for all infra, app-of-apps pattern (Scentbird, Azuro)
- Argo Rollouts [built] — canary deploys, analysis templates, automated rollback (Scentbird)
- Argo Workflows [built] — replaced CronJobs for scheduled workloads (Scentbird)
- GitHub Actions [built] — reusable workflows, custom actions, matrix builds, self-hosted runners (Azuro, OpenQuest)
- GitLab CI (self-hosted) [built] — self-hosted instance, runner administration (Scentbird)
- Deploy Strategies [built] — blue/green, canary, rollback automation, internal tooling (Scentbird 40+ services)

### Security
- HashiCorp Vault [built] — vault-secrets-operator, vault-secrets-webhook, K8s auth (Scentbird, Azuro)
- OIDC / Keycloak / Okta [built] — Keycloak self-hosted, OIDC integrations, Keycloak→Okta migration (Azuro, Scentbird)
- SAST / DAST [used] — integrated in CI/CD, dependency analysis, secrets detection, container scanning
- VPN [built] — OpenVPN, Outline, VLESS/gRPC, split-tunnel by traffic (helped engineers in restricted countries)

### Databases
- PostgreSQL [built] — schema design, Aurora migration, RO replicas, vacuums, partitioning, parameter tuning, pg_stat_statements, PgHero (Scentbird, Azuro, OpenQuest)
- pgBouncer [built] — connection pool reduction under high load (Azuro)
- ClickHouse [built] — CQRS query layer for analytical workloads (Azuro)
- MongoDB / DocumentDB [used] — admin, migration to DocumentDB (Scentbird)
- Redis [used] — caching, distributed locks, pub/sub, queues, rate limiting
- Qdrant [built] — vector database for LLM-era use cases (Scentbird)

### Messaging & Streaming
- RabbitMQ [built] — AMQP event-driven architecture, dead-letter queues, topology design (Azuro — entire inter-service communication)
- Kafka / Redpanda [built] — deploy, topic configuration, consumer groups, retention tuning (Scentbird, Azuro)
- Sequin [built] — Postgres WAL → Kafka CDC streaming (Azuro)

### Blockchain / EVM
- EVM Nodes [built] — archive nodes across 7 chains, HA RPC setup, monitoring (Azuro)
- Custom RPC Proxy [built] — block-lag-aware fallback, not round-robin (Azuro — no market solution fit)
- The Graph [built] — B2B2C subgraph design, self-hosted, pseudo-chain data feed as Hasura alternative (Azuro)
- Hasura [built] — Postgres + ClickHouse connectors (Azuro)
- Solidity [used] — contract reading, integration, debug
- web3.js / ethers.js [used] — contract interactions, event listeners, tx relay, gas error handling
- TON smart contracts [used] — on-chain verifiable quest completion (OpenQuest)

### Backend & Languages
- Node.js [built] — 12+ years, REST API, SSR, WebSocket, blockchain indexers, microservices
- TypeScript [built] — fluent, production use since 2018
- NestJS [built] — microservice architecture, CQRS module, guards & interceptors, AMQP transport (Azuro)
- Java / Spring Boot / Micronaut [knows] — maintained and debugged as SRE at Scentbird, not authored
- GraalVM Native Images [used] — optimised Java microservice startup and memory at Scentbird

### Analytics Stack
- Airflow [built], Airbyte [built], Snowflake [built], RudderStack [built] — analytics pipeline at Scentbird
- dbt [used] — analytics transformation pipelines (Scentbird)

### Feature Toggles
- Unleash [built], GrowthBook [built] — self-hosted feature flag platforms (Scentbird, Azuro)

### Frontend (foundation, 2013–2022)
- React.js [built] — SPAs from zero, SSR with Node.js, e-commerce, web3 dApps, performance optimisation (FCP, TTI)

## Personal

Loves traveling. Doesn't dodge responsibility — production incidents have found him in a bar in Belgrade and in the Tokyo metro. Timezone and vacation don't matter when something is on fire.

## Education

Master's Degree — Automated Control Systems
ITMO University, Saint Petersburg · 2010–2015

## References

People I'm proud to have worked with. Feel free to reach out and ask what it was like working with me.

- Dan Kaizer — CTO @ Azuro · https://www.linkedin.com/in/denkaizer/
- Andrei Rebrov — Ex CTO @ Scentbird · https://www.linkedin.com/in/andrebrov/
- Alexey Zhokhov — Principal Architect @ Scentbird · https://www.linkedin.com/in/donbeave/
- Pavel Pantyukhov — Co-Founder @ Initflow · https://www.linkedin.com/in/pavelpantyukhov/
- Pavel Ivanov — Co-Founder @ OpenQuest (Telegram) · https://www.linkedin.com/in/pavelorso/