Open to opportunities

Alexey Belenkov

@alexeybelenkov

Message

I build production NLP, LLM, ranking, and high-QPS ML systems.

Georgia

Message

What I'm looking for

I'm looking to build production AI and ML systems where I can apply NLP, LLMs, retrieval, ranking, and MLOps to measurable product outcomes, with strong evaluation, reliable operations, and cross-functional collaboration.

At Tinkoff, I build product matching and deduplication systems for a marketplace catalog of more than 3M active SKUs. My pipelines reduced duplicate listings in Electronics & Appliances from 20% to 7.7% and support catalog and search workflows.

Previously at NEAR Protocol, I operated anti-fraud and anti-bot scoring systems serving 8–12M daily requests under 120 ms p95. I combined classical ML, embeddings, Transformers, LLM scoring, RAG, and fallback logic to reduce false positives from 11% to 3.4% at 95% recall.

I enjoy turning ambiguous needs into maintainable AI systems with reproducible evaluation, monitoring, documentation, and reliable deployment workflows.

Experience

Work history, roles, and key accomplishments

Current

Machine Learning Engineer

Current

Tinkoff

Nov 2025 - Present (9 months)

Designed and implemented an offline product-deduplication pipeline in Airflow processing 3M+ active SKUs, reducing duplicate listings from 20% to 7.7% in Electronics & Appliances. Improved a CatBoost product-pair matching scorer and built a fully reproducible experimentation pipeline.

Python Airflow CatBoost Qdrant LLM Git Docker CI CD Monitoring

Education

Degrees, certifications, and relevant coursework

Belarusian State University

Bachelor of Mathematics and System Analysis, Mathematics and System Analysis

Bachelor of Mathematics and System Analysis from the Faculty of Mechanics and Mathematics, with a strong foundation in probability, statistics, algorithms, optimization, data analysis, machine learning, and system analysis.