Gennadiy Moroz

Expert in backend engineering, developing Node.js and TypeScript solutions for high-load environments. Skilled in architecting production systems with robust data integrity and real-time analytics capabilities. Currently overseeing backend operations and engineering standards, ensuring optimal performance across complex, concurrent requests.

• Analytics at scale, Designed ClickHouse-backed event aggregation with buffered ingestion, slice-based reporting, and Grafana integration for both system and business metrics.
• Browser-side instrumentation, JS reverse engineering and undocumented HTTP API analysis to integrate and debug third-party systems where no documentation existed.
• Built and shipped applications, including a native Mobile app SDKs.
• Active participant in hackathons and application-development competitions.

• Analyzed solution performance and revenue effectiveness end-to-end — connecting technical metrics (latency, throughput, fill) to business outcomes (yield, revenue per request) to decide what actually moves the numbers.
• Designed and ran A/B tests on production traffic: gathered and compared report data across variants and made ship/rollback decisions based on measured revenue and performance impact rather than intuition.
• Built feature-flag systems for runtime enable/disable of features, including hot-reloading env-variable vs HTTP-API toggling strategies — allowing safe rollout and instant rollback without redeploys.
• Architected modular, pluggable systems (optional feature modules, factory patterns) so features could be composed, isolated, and measured independently.
• Implemented frontend cache-busting (versioned/hashed build artifacts) and CDN/edge cache invalidation on release, ensuring clients reliably receive new versions while controlling Core Web Vitals impact.

• Production rewrite for performance: Re-architected a core ad-serving server from Node.js to Go, improving throughput and resource efficiency under high load.
• Document-store experience: MongoDB from MEAN-stack high-load systems through to microservice caching layers; Redis as a primary caching and state tool throughout.
• Distributed & event-driven: Message queues, async job pipelines with retries and state management, RPC inter-service sync, and horizontal scaling across clusters.

• English (professional)
• Russian (native)
• Ukrainian

• HeroVideo.io (production product), Engineered AI-driven video generation pipelines — orchestrating generative video and image models, content parsing, and prompt/context engineering into a production system serving real users.
• LLM integration, Built production features on both commercial APIs and self-hosted open-source models; designed prompt and context-engineering strategies as repeatable engineering practice rather than ad-hoc prompting.
• Retrieval / RAG, Built a bilingual meeting-intelligence pipeline with hybrid retrieval (BM25 + dense embeddings) exposed through a custom MCP server.
• Data extraction & summarization, Web-site and feed parsing, normalization, and LLM-based summarization feeding downstream product and analytics workflows.
• Self-hosted inference, Architected multi-GPU serving infrastructure for running open-source models in-house, including hardware-level diagnostics and monitoring.