Improving efficiency of recommendation system used by News and Feeds.

- Performed migration of default experiment settings storagefrom old one-dimensional a/b testing system to the newmultidimensional one. It allows to conduct 10 times moresimultaneous experiments compared to the previous one. Ialso had to write code in such a way, that the new systemcould work alongside the old one for the time of thetransition.- Implemented several production features in collaborationwith ads machine learning team, e.g. integrated newprediction models with ads engine. That increased totalrevenue of our machine learning team by several percents.- Performed onboarding of 3 software engineers (one ofthem is senior software engineer), two of which continued towork with our team. Mentored them for some time.- My responsibilities also included but were not limited to regularads engine debug and a/b testing system maintenanceduties.

- Implemented migration of user identification servicemessage format from JSON to Protobuf. It reduced networkusage by 40%+ and total ads engine CPU usage by 3%+.

- Researched and developed branch-and-bound fraudprediction algorithm for fraud analysis team.- Migrated several antifraud regular processes from oldscheduler to the new one.