Developed a streaming join optimization for the ytsaurus/ytsaurus repository, introducing the ReverseBlockJoin feature to enhance LEFT JOIN operations in BlockHashJoin. This approach enabled building the left side of a join while streaming the right side, which improved flexibility and efficiency when processing large datasets and real-time streams. The work focused on reducing latency and resource usage in analytics workloads by leveraging advanced algorithm design and data processing techniques. Implemented entirely in C++, the solution demonstrated a strong grasp of streaming algorithms and join optimization within a core data platform, delivering a targeted feature without reported major bugs during the development period.