Sergei Elnikov Email & Phone Number

Belgrade, Serbia

Designed, developed, and maintained robust ML infrastructure to support high-load operations within a large-scale delivery service environment.• Implemented distributed service for real-time geographical data processing. Leveraged the h3 library, Kafka-like message broker, and Redis functionalities to optimize performance. Successfully resolved complex performance challenges inherent in distributed systems, including replication lag, message ordering, and sharding.• Engineered a service for asynchronous monitoring and controlling of demand/supply balance using statistical features, ML predictions, and graph algorithms. Ensured comprehensive logging coverage for ML and analytical purposes. Supported efficient infrastructure for shadow deployment of new algorithms for rapid hypothesis testing.• Established extensive observability coverage to facilitate seamless system maintenance. Employing tools such as OpenTracing, Grafana, and proprietary observability solutions to ensure efficient monitoring and troubleshooting.• Implemented services using C++ and an asynchronous framework for web services (userver), employing multithreaded programming techniques for sustained high performance. Additionally, utilized pandas alongside YQL (SQL dialect) for preliminary performance and data analysis

St Petersburg, St Petersburg City, Russia

Contributed to enhancing and expanding the capabilities of a platform dedicated to storing and processing hundreds of terabytes of data for one of the country's largest online marketplaces.Focused on ensuring data integrity and performance optimization in a high-load environment, where thousands of events per second were processed, guaranteeing minimal disruption even in unforeseen scenarios.• Engineered a fault-tolerant distributed pub-sub service using C++, alongside an in-house coroutine engine, Protocol Buffers, and YTSaurus OLTP solution.• Developed batch data processing pipelines utilizing the YTSaurus MapReduce C++ API, enabling efficient handling of large datasets.• Leveraged tools like valgrind and perf to analyze and enhance the performance of intricate high-load services.• Implemented user-friendly features such as automated dashboard generation, alert systems, and a tool for selective data recovery from backups, enhancing both platform usability and developer productivity.

St Petersburg, St Petersburg City, Russia

• Implemented PySpark as a new solution for the data processing pipeline, replacing the previous in-house Scalding-like DSL solution, thereby enhancing efficiency and scalability• Post-implementation of PySpark, analyzed and optimized MapReduce pipelines to improve speed and cluster resource utilization

St Petersburg City, Russia

Conducted practical classes with code-review for first-year students on development tools.

St Petersburg, St Petersburg City, Russia

• Developed algorithms for 5G base station data processing using C++• Used Matlab to write integration tests