Evgenii Kocherga

Azerion - Media Platform with core business around advertisements. As a Data Scientist, my responsibility to maintain different models around pairing user/publisher with campaign/advertiser.

Among projects:

1. CTR prediction for comparing CPC and CPM bids

2. Advertiser response prediction

3. Productionalising ML models with Scala, H2O and AWS

Petal - General Search Engine of the Huawei.

As part of the team, I had two main responsibilities:

1. Generate new features for ranking model as documented and wrapped with tests production code.

2. Quality Control of assessors, marking quality of query-document pairs.

As the most proud result - highly modified adaptable variation of Dawid-Skene algorithm as the instrument to convert unreliable answers from crowd-platform into measurable quality of models and assessors.

Develop and tuning of ML models, based on neural nets.

The company itself - b2b for mobile game developers, provided variety of services to keep track on efficiency of the game. My part was to create a model, suggesting some possible in-game pop-ups to encourage in-game purchases. As it's hard to maintain every model separately, and each game has its own unique set of actions, purchases, currencies, statuses and overall flow of gaming experience, I had to create sort of Fabric for NN models, create a proper validation and testing over multiple datasets and maintaince/reporting dashboards about it.