Sergey Goshkov

- Research and creation of a new machine learning model for no-text ranking with new features: +0.3% of the total number of buyers, +0.3% of the total number of new buyers in production. - Search suggest quality improvement by 1%: the creation of geo-tips, navigation deeplinks, category filters. - Leading the addition of catalogs of trucks and special equipment, developing of infrastructure for AB testing for this. Creating a new tire size match: +1.5% in new tire buyers, +3% in new contacts.

- Correction of dozens of cases of errors in recognizing the boundaries of texts, markup, and other entities. - Development of an MVP recognition service for Business Card Reader from scratch.

- Development universal e2e quality metric for voice assistant, using metric to accept new scenarios. - Development and analytics 5 new scenarios for voice control of smart home and smart speaker. - Creation of regular data labeling processes. This is used to train all intent detection models.

- Creation of offline search quality metrics, a tool for assessing the relevance using crowdsourcing. - Conducting experiments on the use of gradient boosting in traffic forecasting.