Matvey Tolkunov

DevOps Engineer · Ecom.tech (Samokat)

Internal ML platform serving hundreds of active users — data scientists and ML engineers.

• Solely operate and evolve 6 instances of JupyterHub — a service with hundreds of active users.
• Built a custom JupyterHub–Spark integration for interactive work with the cluster — sped up the model training cycle in production.
• Built platform monitoring from scratch: service error metrics, GPU, CPU, memory and disk utilization down to nodes, pods, roles and users; a dozen dashboards in Grafana and alerts to the corporate channel — monitoring as a service.
• Diagnosed and fixed hangs in Airflow DAGs training models on order data — restored the stability of promo and product analytics pipelines.
• Introduced linting across all Airflow repositories — cut DAG errors and leveled up code quality across teams on more than 10,000 DAGs.
• Implemented SSO: Keycloak integration with JupyterHub.
• Deploy and operate LLM services for internal teams: Triton, plus Langfuse for tracing.
• Developed a set of CLI utilities that make platform users' day-to-day work in production easier.
• On-call in the incident rotation: every incident during my shifts was driven to resolution.

DevOps Engineer · RSHB-Intech (Rosselkhozbank)

Infrastructure of the RAISA ML platform: CI/CD, Kubernetes, monitoring, services for data science teams.

• Built CI/CD pipelines in GitLab CI — from image builds to automated deploys to Kubernetes; release time went from 3 days down to 1 hour.
• Customized and deployed MLflow: OAuth2, PostgreSQL, artifacts in S3 (MinIO), load balancing via ingress-nginx — plus a pipeline that deploys the whole stack automatically.
• Deployed the monitoring and logging stack: Prometheus, Grafana, VictoriaMetrics, OpenSearch, Fluent Bit.
• Migrated JupyterHub user environments from Docker Swarm to Kubernetes — about 1,000 users, with zero data loss.
• Automated uploading large LLM models to Nexus Sonatype.
• Configured HAProxy and Nginx: domain-name routing and load balancing.

DevOps Engineer · Clean Data

Rolling out and supporting the company's product on customer infrastructure.

• Deployed, configured and supported the company's software on customer premises.
• Developed CI/CD for product modules (TeamCity, GitLab).
• Ran data migrations and initial data loads, configured deduplication and normalization.
• Troubleshot Java applications across dev/test/prod environments.
• Wrote a complete set of product administration guides.

Software Engineer · RSSU Technopark

• Developed modules for service robots on ROS (Python).
• Built a series of three-channel video recorders on Raspberry Pi (Debian).
• Designed a smart-home system on Raspberry Pi 4B.

Systems Engineer · SKTB PR

• Developed the GUI and menu navigation for an industrial cable robot inspecting defects inside oil pipelines.
• Set up Debian-based workstations.