Nikolay Kotlyarov

Data Scientist

VISA

Global payments technology company and card network.

Built transaction analytics, predictive features, classification and fraud models, and data-quality systems; reported revenue, adoption, model-performance, and operational improvements.

• Created tokenization analytics on 10M+ payment transactions, identifying optimization opportunities that increased net revenue by 8% and improved token adoption rates by 15%. • Performed statistical analysis and feature engineering on datasets containing 100M+ transaction records, developing predictive features that improved model performance by 11%. • Developed PyTorch-based machine learning models for merchant classification and fraud detection, achieving 95%+ classification accuracy and reducing false positives by 19%.

Data Scientist

Axenix

Management and technology consulting company.

Developed and deployed LLM and NLP models and infrastructure, improving developer productivity, reducing GPU memory use, and increasing inference throughput.

• Developed and deployed a Transformer-based code generation model using PyTorch and Hugging Face, improving developer productivity by 25% and reducing code generation latency in production environments. • Designed a memory-efficient RLHF training framework for large language models, reducing GPU memory consumption by 33% and enabling more scalable model training. • Researched and implemented parameter-efficient and full-model fine-tuning techniques for large language models, achieving double-digit improvements across text classification and information extraction benchmarks.

ML/Data Engineer

QAZDEVELOP

Built a document-signing platform and backend services, led a microservices migration, and improved CI/CD deployment time by 70%.

• Created a digital document-signing platform for counterparties, reducing contract turnaround time by 50% and eliminating manual approval bottlenecks. • Led the migration of a monolithic application to a microservices architecture, increasing development velocity by 30% and improving system scalability. • Designed and implemented RESTful APIs and backend services handling 100K+ daily requests, improving system performance and integration capabilities. • Built and optimized CI/CD pipelines using Git, Docker, and automated testing, reducing deployment time by 70%.

Master's, Data Science

Kazakh-British Technical University

Bachelor's, Computer Science

Karaganda Technical University