Arsen Ozhetov

Optimized analytics by redesigning file-reading logic, introducing shared caching, and removing redundant multiprocessing, reducing duplicate SharePoint XLSX reads by 90% and cutting job startup computation by 5x. Re-architected cron-based ingestion into a WebSocket-driven non-blocking service to reduce data latency from minutes to seconds.

Improved a data processing pipeline up to 2x by consolidating scheduled background tasks into the main flow, eliminating cross-service polling, reducing duplication, and enabling real-time processing. Increased Load Balancer unit test coverage to 80% and implemented unified log streaming for faster debugging.

Enhanced AI model accuracy by 15% by writing 500+ lines of SQL scripts and automating daily AI log collection with cron jobs to improve training system performance. Built features for Yandex Search and Alice, including regionalization and an error alert system using Java (Spring Boot) and Python (Flask).

Optimized a store leftover calculation service using Python, FastAPI, and SQLAlchemy with hashmap and prefix-sum algorithms, achieving a 750% performance improvement. Reduced warehouse goods collection time by 20% via an optimization routing service and improved scalability using JMeter load tests simulating up to 10,000 users.

Developed high-performance APIs with a 4-layer architecture using Java and Spring Boot, integrating MySQL and Redis caching to improve response time by 80%.