Maksim Petrov

• ● Developed a logistic regression scoring model for credit card applicants using SAS feature engineering and Python data analysis, increasing Gini by 10 p.p.

• ● Built an XGBoost classification model (Gini 64%) that helped banks predict applicant behavioral profiles, improving risk understanding across different customer cohorts.

• ● Quarterly monitoring of production models via dashboards tracking Gini, gain charts, score distributions, PSI, and variable stability.

• ● Optimized internal data processing workflows on CentOS servers by automating SAS jobs and integrating Python scripts, reducing runtime by 5 hours.

• ● Collaborated with the top 4 CIS insurance companies that helped to analyze car insurance market data and build a consistent model for fraud prediction (Gini 49%).

• ● Reduced fraudulent participation in trading tournaments by 90% through designing and implementing a clustering-based detection algorithm.

• ● Developed a fraud-detection service for a team of 20 international affiliate managers, cutting manual review time from half a day to 5 minutes and significantly lowering resource usage.

• ● Built an internal Tableau dashboard that reduced user verification requests response time by 80% by consolidating all required information into a single view.

• ● Conducted an A/B test that demonstrated the inefficiency of the existing scoring model, leading to its replacement with a rule-based approach and avoiding unnecessary model complexity.

• ● Enhanced Probability of Default (PD) business model quality for IFRS 9 compliance by developing a credit card customer classification algorithm (Gini 51 %).

• ● Improved overall PD model performance by designing and implementing a mortgage scoring model (Gini 68%) using advanced statistical methods.

• ● Built a comprehensive monitoring system, including automated reports and dashboards to track PD model performance and data quality across portfolios.