Vladislav Khorunzhiy

Go backend for a Bitcoin mining pool serving about 20,000 clients across Stratum relay, hashrate routing, and share processing.

• Rewrote the Stratum proxy runtime and separated connection handling from share processing; a relay pod handles up to 500 RPS on 2 CPU cores.
• Implemented seamless hashrate switching across pools without dropping miner TCP sessions, supporting up to 40,000 ASIC connections at peak.
• Added JobID-aware upstream handoff: establish the new pool connection before closing the old one; during the transition, forward only submit messages for jobs from the previous pool.
• Handled up to 25,000 Redis write operations per second.
• Replaced blocking prefetch with PeekWithTimeout, enabling an immediate response to mining.subscribe.
• Built EWMA-based hashrate distribution across upstream pools to smooth short spikes and maintain target share rates.
• Traced a false hashrate drop across the Redis → collector → Grafana path; a one-hour Redis collector stall caused internal accounting to miss shares accepted upstream.
• Fixed two bugs in the reference stratum-mining/stratum Stratum V2 implementation: a panic on invalid BIP141 coinbase data and a stale channel ID after failed group addition; one patch shipped in a major crate release.
• Unified rate feeds from five external pools behind a shared interface for profitability comparisons.

Go · PostgreSQL · Redis · Kafka · ClickHouse · Prometheus · Grafana · Loki · Mimir · Tempo · Docker · Kubernetes · Linux · Git · REST API

Go backend services for telemetry across a fleet of 30,000+ vehicles, handling packet intake, vehicle state and status, alerts, and product features.

• Migrated telemetry processing to gRPC and Kafka as vehicle data volumes grew.
• Built a Go car emulator that sends test packets through the production telemetry path, cutting manual scenario setup from about 30 minutes to seconds.
• Integrated the emulator into automated pre-release telemetry checks to validate new fields alongside existing rules.
• Added five failure scenarios: invalid coordinates, value jumps, conflicting sensor readings, connection loss, and duplicate submissions.
• Tuned Kafka consumer concurrency for telemetry ingestion, reducing p95 processing latency from 1.8 seconds to 420 ms.
• Added schema-compatibility checks to GitLab CI to catch incompatible CAN fields before release.
• Automated CAN-provider packet checks, reducing malformed packets reaching production by 70%.
• Added deterministic replay of recorded vehicle packets to reproduce field failures in under two minutes.
• Built packet-lag and sensor-anomaly dashboards, reducing median triage time from 45 to 12 minutes.

Go · gRPC · Kafka · PostgreSQL · Redis · ClickHouse · Docker · Kubernetes · GitLab CI · REST API

Go services for B2C internet banking and payment-system integrations.

• Built an incoming-payments service with idempotent operation processing.
• Reduced PostgreSQL load by 30% with Redis and asynchronous processing through Kafka.
• Optimized the mobile GraphQL gateway with batching, reducing p95 response time from 310 ms to 120 ms.
• Partitioned Kafka payment consumers by account key, raising throughput from 800 to 2,400 messages per second while preserving order.
• Added idempotency keys to partner callbacks, reducing duplicate payment postings by 99.9%.
• Built a payment-reconciliation state machine, cutting manual exception work by 65%.
• Moved partner adapters to a shared gRPC contract, reducing integration lead time from two weeks to five days.
• Added payment latency and failure dashboards, cutting incident diagnosis from 40 to 15 minutes.
• Added payment-flow regression tests to GitLab CI, cutting release verification time by 50%.

Go · TypeScript · REST · gRPC · GraphQL · WebSocket · PostgreSQL · Redis · Kafka · GitLab CI