Internet advertisement company with 10k+ clients worldwide and 6B+ ad impressions daily.R&D team, responsible for all ML and analytics in the company.Increased company profit by 10% (confirmed in A/B) with a new advertisement rotation algorithm:- Designed algorithm for contextual multi-armed bandit problem;- Built statistical LTV model for delayed feedback campaigns;- Implemented custom hierarchical clustering for grouping similar websites;- Designed cross-format advertisement rotation approaches;- Presented the project to business departments and C-level managers.Designed Machine Learning solutions:- Prevented 100k$ in annual losses with multivariate time series-based anomaly monitoring;- Doubled traffic fraud detection with a semi-supervised approach;- Speeded up A/B tests 3-5 times by reducing variance with the CUPED covariate model.Participated in team development:- Hired and onboarded 3 data scientists/analysts;- Reviewed and improved colleagues’ solutions on insights tasks;- Implemented analytics review, code review, unit testing and knowledge exchange.

Bank with 2M+ individual clients, 100k+ business clients and a wide range of credit products.Credit risk modelling department.Increased scoring AUCROC on 3-5% for several credit products:- Applied logit and glmtree for application and 3rd party data models;- Uplifted credit history-based model with gradient boosting model;- Built a new ensembled model based on behavioural data.Created a library for specific modelling tasks:- Unified and automated development and reporting processes;- Made the analysis and model development process 3 times faster;- Increased model’s quality with automated feature engineering and feature selection modules.