Commodities trading platform

Worked at the Core team. Developed high frequency trading platform engine:

• Custom solutions instead of libraries for reducing latency: network protocols, lockfree data structures

• Custom low-latency logging and tracing systems

• High-throughput async networking library

• Framework for building exchange pipelines with compile-time binding and validation on C++20 using boost::mp11 and boost::hana

Worked at localization and mapping team. Mainly developed the process of lidar map creation:

• Implemented two versions of dynamic objects removal from the map: using perception detections and several self-written statistical methods of filtration

• Improved hyper-graph optimization algorithm, which allowed to improve map accuracy from 50 cm to 1-5 cm

• Parallelized the process of map creation and graph optimization, which led to speed up algorithm several times. Implemented dynamic loading of the lidar map to the localization component which reduced RAM consumption significantly

• Implemented camera and lidar data fusion (coloring of lidar data) on C++, which led to 10x speed up compare to old python algorithm and allowed to perform the fusion in realtime for better objects detection

• Implemented algorithm for merging of several lidar maps into one, which allowed to create maps from the data that doesn't fit on the vehicle disk

• Implemented several CLI tools for HDMap team for the convenient hdmap creation process

Worked at backend development team of drivers app and fleet management tool products.

• Developed driver and cars profile cards backend, API for extraction of this data by the internal and external services

• Developed service for collecting statistics about drivers working time and vehicle vehicles using. It helped taxi stations to use these metrics for the processes improvements

• Implemented a new version of the infrastructure service for distribution of cars according to tariffs. The old service had complicated logic and was unclear in use for managers. The new version kept backwards compatible with the old version but had more clear logic.

Languages: C++, Python, C#

Databases: PostgreSQL, MongoDB, Redis, ClickHouse