Core infrastructure
Kubernetes, GCP, Azure, Linux, Docker, PostgreSQL, multi-region and hybrid-cloud production systems
Lead Platform Engineer · AI Infrastructure & Reliability
Bekhruz
Kasymov
Lead Platform Engineer with 5+ years of experience building reliable infrastructure for international AI, HealthTech, TravelTech, InsurTech, and banking products. Operated 8 Kubernetes clusters and 30+ microservices across five continents, supporting products used by 333K+ active users and approximately 400 clinics while reducing release time by 80%, cloud spend by 40%, and incident response time by 50%.
333K+ active usersConfidential AI platform
~400 dental clinicsDiagnocat global platform
~6,000 customersWeGoTrip during tenure
8 Kubernetes clustersGCP and Yandex Cloud
80% faster releases4–6 hours to under 1 hour
40% lower cloud spendCleanup and spot autoscaling
Technical profile
Skills mapped to production systems
Infrastructure & delivery
Terraform, Terragrunt, Ansible, Helm, ArgoCD, GitLab CI/CD, Jenkins, GitHub Actions, GitOps
Reliability & security
Prometheus, Grafana, Loki, Alertmanager, Sentry, Trivy, SAST/DAST, RBAC, ingress and certificate management
AI platforms & data
vLLM, Ollama, ClearML, KubeRay, GPU workloads, Python, MongoDB, Redis, Kafka, RabbitMQ
Professional experience
Engineering reliable delivery
Confidential AI Platform
Lead Platform Engineer
- Own a hybrid Kubernetes platform spanning self-hosted development environments, Azure AKS production, and Nebius GPU infrastructure for AI workloads.
- Operate production infrastructure serving 333K+ active users across a large portfolio of AI generation and productivity workflows.
- Completed the migration from Bitbucket to self-hosted GitLab, including repository moves, backup retention, protected reviews, CI/CD rewiring, and GitLab Runner deployment on Azure Kubernetes.
- Implemented highly available PostgreSQL in Azure after a managed-database outage caused mass 5xx responses and authentication failure; strengthened database monitoring and environment maintenance workflows.
- Built AIOps workflows with HolmesGPT, MCP-connected data analysis, and automated service-error reporting to detect API/ML anomalies and shorten investigation loops.
- Hardened production ingress against spoofed client-IP headers that could bypass authentication rate limits or block arbitrary users; evaluated Azure WAF controls and trusted-proxy boundaries.
- Resolved recurring GPU/NVML and AI-generation incidents, added health probes and resource controls, and operated observability across Prometheus, Grafana, Loki, Sentry, Graylog, and Fluent Bit.
- Built a repeatable Azure proxy layer for external AI providers, solving regional routing, TLS/SNI, certificate, large-payload, WebSocket, IPv6, and timeout edge cases.
Diagnocat
Senior DevOps Engineer
- Managed three outsourced DevOps engineers and coordinated infrastructure strategy across frontend, backend, DevOps, and MLOps teams.
- Architected and operated eight Kubernetes clusters on GCP and Yandex Cloud, running 30+ microservices for approximately 400 dental clinics across five continents with high availability and disaster recovery.
- Reduced production release time from 4–6 hours to under one hour by automating tests and deployment workflows—an 80% improvement in time-to-production.
- Cut cloud infrastructure spend by 40% through resource cleanup and autoscaled spot-based GitLab Runners.
- Built 30+ GitLab CI/CD pipelines with Trivy, SAST/DAST, Terraform, Ansible, Helm, and ArgoCD-based GitOps controls.
- Deployed four or more production AI models—including Whisper and open-source LLMs—with KubeRay, Ollama, vLLM, ClearML, and GPU autoscaling; reduced external AI API costs by 4×.
- Implemented Prometheus, Grafana, Loki, and Alertmanager across the platform and operated Istio, ingress, and certificate layers, reducing incident response time by 50%.
WeGoTrip
Senior DevOps Engineer
- Owned infrastructure for eight or more services across four cloud providers, supporting 10+ engineers and a platform serving approximately 6,000 customers with zero unplanned downtime.
- Built and maintained 15+ CI/CD pipelines for frontend and backend services, improving the reliability and repeatability of releases.
- Reduced cloud infrastructure costs by up to 70% through resource right-sizing, provider selection, and Terraform/Terragrunt automation.
- Strengthened disaster recovery and shared production knowledge across product engineering teams.
Qantev
DevOps Engineer · Contract
- Delivered an AWS-to-VK Cloud migration in four months while maintaining delivery continuity for an international product team.
- Supported cloud delivery for an international AI health-insurance platform trusted by 20+ insurer clients worldwide.
- Built Azure infrastructure with Terraform and Terragrunt for an AI-driven health insurance platform.
- Optimized delivery workflows and reduced build time by 30%.
Sberbank
DevOps Engineer · Software Engineering Intern
- Built and maintained eight or more Jenkins delivery pipelines using GitLab, SonarQube, Python, and Shell automation.
- Trained two engineering teams—more than eight engineers—on delivery workflows and operational practices.
- Reduced deployment errors by 30% through repeatable automation and improved quality controls.
Selected additional engagements
Focused work beyond core roles
WeChargers
Infrastructure consulting
Production reliability and infrastructure work for a power-bank rental platform in Georgia: Docker-based service operations, observability, deployment safety, and investigation of database and application-performance incidents.
WeGoTrip
Post-employment consulting · Project-based
Returned after the full-time role for a separate infrastructure engagement, providing delivery continuity and applying deep knowledge of the company’s multi-cloud production environment.
Volunteer experience
Service across law, technology, and community
Sep 2015 — Dec 2017 · Human Rights
Legal Consultant
Kutafin Moscow State Law University (MSAL)
Provided pro bono legal assistance to socially vulnerable communities.
Sep 2017 — Jan 2018 · Science and Technology
Assistant to an Associate Director
Created, introduced, maintained, and processed technical documentation for the organization.
Sep 2021 — Present · Science and Technology
Technical Consultant
Gareni / Pamix / Tabler
Advise colleagues on programming languages, databases, automation, and optimization.
May 2023 · Social Services
Distributor
Supported a one-month social-services distribution initiative in Paris.
Jun 2023 · Arts and Culture
Administrative Assistant
Provided short-term administrative support at the Paris startup campus.
Jun 2023 · Science and Technology
Volunteer Staff
Supported on-site operations at the Viva Technology event in Paris.
Selected complex systems work
Problems solved in production
01
Diagnocat · Scale · Reliability
Global Kubernetes platform
Architected and maintained eight Kubernetes clusters across GCP and Yandex Cloud, supporting more than 30 microservices in five continental regions with high-availability and disaster-recovery requirements.
Outcome: release cycle reduced from 4–6 hours to under one hour; incident response time reduced by 50%.
02
Confidential AI Platform · PostgreSQL · Azure
Production resilience after DB outage
Responded to an Azure database outage that caused mass 5xx responses and broken authentication, then delivered a highly available PostgreSQL production setup and improved database visibility and maintenance controls.
Closed production work: HA PostgreSQL, replica-load analysis, Grafana datasource and environment cronjob alignment.
03
MLOps · GPU · AIOps
AI platform operations
Combined model-serving operations with automated incident analysis: production LLM inference at Diagnocat; GPU/NVML recovery, readiness controls, HolmesGPT, and MCP-connected service anomaly analysis at a confidential AI platform.
Outcome at Diagnocat: four or more models in production and 4× lower external AI API costs.
04
Confidential AI Platform · Security · Delivery
Secure developer platform
Migrated delivery workflows to GitLab, hardened ingress client-IP trust, introduced managed credential sharing and database RBAC, and investigated WAF protection for the production edge.
Security work covered the full path from source control and secrets to ingress, rate limits, database access, and production networking.
05
Selected product · Full stack · AI
FailUp — AI product built end to end
Designed and built a production personal-growth platform across React, Node.js, Flutter, MongoDB, Redis/BullMQ, Qdrant, and OpenAI APIs, including semantic search, asynchronous AI workflows, encryption, and data export.
Demonstrates product ownership beyond infrastructure: application architecture, mobile and web delivery, AI pipelines, privacy controls, and self-hosted K3s operations.
Latest writing
Ideas behind dependable systems
Systems Thinking
Stability Starts with Discipline
Why real stability comes from personal discipline, responsibility, and the ability to keep acting when circumstances change.
Software Engineering
How OCaml Changed the Way I Think About Code
What two difficult OCaml projects at École 42 taught me about expressions, types, state, and becoming a more disciplined engineer.
AI Infrastructure
How to Implement AI in DevOps in One Evening
A practical, low-cost architecture connecting observability, source control, an LLM, and team chat to answer infrastructure questions automatically.
Education
Systems, software engineering, and law
2022 — 2025
42 Paris
RNCP Level 7 · Expert in IT Architecture — Information Systems & Networks
Project-based computer science training centered on independent problem solving, peer learning, software engineering, and systems thinking.
2019 — 2022
School 21 by Sber
Software Engineering Program
Peer-to-peer, project-based training in software engineering, systems programming, and collaborative development.
Graduated 2018
Kutafin Moscow State Law University
Law
Legal education supporting structured analysis, risk assessment, governance, and compliance-minded engineering.
Languages
Russian · Tajik · English
Next opportunity
Ready to make high-scale, cross-layer systems understandable and dependable.
I bring a builder’s mindset, production ownership, and a bias toward clear, automated, maintainable infrastructure.