Golang Developer

Organised a multi-sharded ClickHouse cluster (16 machines) and real-time analytics for market data and execution reports. Peak load scaled from 1M → 55 million MD ticks per minute. Real-time market data processing using ClickHouse Materialized Views. Created real-time monitoring and alerting system (Grafana + ClickHouse + Shared Memory). Market data roundtrip: 100 µs (ultra low latency). Go latency improvement: heap and GC optimisations that eliminated 10 ms latency spikes. C++ code refactoring and latency improvement for market data on the critical path.

Built and led an affiliate team of 4 engineers from scratch. Designed and managed technical architecture of 20+ services (multi-region) for affiliate advertisement tracking. Achieved SaaS cost reduction of >$1M USD/year. Managed high-load tracking system: 2B events/month, 200K orders/month. Optimised business costs by >$3.5M USD/year via fairer financial flows in advertisement. Provided instruments to increase GMV by more than 1.5×.

Ported ICU (International Components for Unicode) library to Go. Reduced p90 latency per call from 10 ms → 100 µs. Solely developed 'favourite category' buyer discount feature: financial flows, UI, NoSQL storage, load balancing, caching. Implemented internal loyalty bonus system serving 200+ company members with e-commerce currency and feedback collection.

Maintained high-load buyer behaviour tracking system at 1K+ RPS across the full website. Doubled data capacity on a 10 TB+ ClickHouse cluster via intelligent partitioning and sorting. Built session mechanics service with Prometheus metrics; cut analyst query time from 10 h → 1 h.

Maintained reverse search index for the full Yandex image search (~10 PB average index size). Implemented stress-load optimisation: deprioritised low-click results in favour of popular ones. Reduced index build time by 10% through build process optimisation.

Upgraded protocol for Chicago ↔ New York microwave channel: 2× latency reduction vs fibre optic. Built automatic latency-checking logic using Valgrind and Cachegrind. Latency monitoring and improvement across all financial strategies for Math & Research teams. Network issue analysis using raw TCP dumps and Q+KDB.