miruzavin

Designed and implemented an end-to-end observability platform for SIP and WebRTC services within Yandex real-time communication platform. Built the full telemetry pipeline from traffic capture and trace generation to storage and analysis in ClickHouse. Deployed and optimized the ClickHouse cluster handling 1 TB of telemetry data ingestion per day. Developed a fault-tolerant, multi-region call recording system for Yandex’s internal telephony platform, supporting 4,000+ concurrent sessions. Transitioned from stateful architecture to ephemeral local storage with S3 persistence, eliminating call losses during host failures. Improved deployment efficiency, reducing rollout time per location (10 pods) from 10 hours to 10 minutes, enabling faster service expansion. Designed and implemented a flexible call routing scheme based on internal short numbers, improving user experience by enabling customizable routing scenarios for employees.

Developed a Terraform provider module that removed unnecessary noise during the Terraform plan stage. Rewrote Refactored the system to support dynamic feature enabling and disabling, which improved system stability, as well as its flexibility by allowing you to transfer 30% of initializing settings.

Developed a module that allows to replicate user files. This increased the fault tolerance of the system and increased download speed. Developed a module for collecting the status of nodes. It improves team understanding of the network state and enables faster response to network failures.