Yerkebulan Salimgereyev

Senior Data Scientist

Kaspi Bank

Kazakhstan-based bank and fintech company.

Built real-time marketplace ranking and recommendation systems, including a purchasing-power model that increased AOV by 10% without reducing conversion.

• Architected and productionized a high-load, low-latency real-time product re-ranking service with ~30 ms p99 SLA; ensured reliability and scalability via comprehensive load and stress testing using Locust. • Designed and deployed a large-scale recommender system for a 15M-user marketplace, combining collaborative filtering with feature-based ranking; directly optimized for key business metrics (CTR, CR, AOV), driving measurable uplift across users, sessions, and events.

Senior Data Scientist

Halyk Bank

A major commercial bank in Kazakhstan.

Developed banking recommendation, uplift-modeling, and client-monitoring systems, with reported conversion uplifts of up to 30% and 50%.

• Designed and implemented an end-to-end pipeline for detecting negative mentions about clients: data ingestion via search APIs (Google CSE), text classification, and LLM-based summarization; enabled automated filtering and concise extraction of actionable insights. • Optimized company name matching algorithms, reducing false positives by 54% with zero recall degradation, significantly improving data quality for downstream systems. • Developed a multi-treatment uplift modeling framework to optimize incentives (purchase categories × bonus sizes), achieving up to +30% conversion uplift.

Data Scientist

Kaspi Bank

Kazakhstan-based bank and fintech company.

Productionized call prioritization, outreach, and volume-forecasting models, and tested data quality for downstream analytics.

• Developed and productionized a client call prioritization model, identifying high-value phone numbers based on historical call success patterns, improving outreach efficiency and client engagement. • Built a targeted call outreach model using stacking (integrating debt collection signals for overdue loans), improving client selection quality and enabling prediction of changes in call responsiveness (improvement vs deterioration). • Designed a predictive model for daily call volume forecasting, providing actionable guidance for dial-up planning and resource allocation.

Data Miner

National Center for Road Asset Quality

Annotated and prepared image data for neural network training.

• Engaged in data manipulation, specifically with images, involving annotation and meticulous preparation for neural network training.

Bachelor, Electrical and Computer Engineering

Nazarbayev University