Data Scientist
/ CV Engineer

I build production-grade ML systems with a focus on computer vision and end-to-end pipelines: from data and training to optimization and deployment.

Python • PyTorch Computer Vision FastAPI • Docker • Linux Optimization & Deployment

About

Short and hiring-friendly

Data Scientist with 2+ years of experience in fintech, delivering ML solutions with measurable business impact. Built credit scoring, behavioral and risk models, recommendation systems, and NLP classifiers, covering the full ML lifecycle from feature engineering to deployment and monitoring.

Core skills

Compact and scannable
Python PyTorch CatBoost scikit-learn Computer Vision NLP (BERT) SQL Airflow MLflow Spark Docker Linux

Experience

Impact-focused highlights
Middle Computer Vision Engineer — Google
Present
  • Working on computer vision systems and production-grade ML pipelines (details available on request).
Middle Data Scientist — Microfinance Institution
Feb 2024 — Oct 2025
  • Developed a CatBoost credit scoring system: default rate −7 p.p., approval rate +11%, > ₽3M annual profit uplift.
  • Built a cascade creditworthiness model: inference time −30%, operating costs −₽5M/year while maintaining accuracy.
  • Automated complaint analysis with a BERT classifier: support workload −25%, NPS +7 points.
  • Designed risk model for high-debt clients: NPL in the segment −9%.
Junior Data Scientist — Technological University
Sep 2023 — Jan 2024
  • Built a hybrid recommender for a digital library: recommendation accuracy +10%, conversion +15%.
  • Developed a real-time object detection + tracking module on a single-board computer: stable 40 FPS.
  • Implemented an ML pipeline for cybersecurity incident classification: 94.2% detection accuracy, processing time −18%.

Selected projects

No external links

Computer Vision Recognition Service

Production inference API with robust logging, monitoring, and optimized latency.

PyTorchFastAPIDockerLinux

Pan-Tilt Camera Tracking System

Real-time detection + tracking driving a pan-tilt camera for stable target lock and smooth motion control.

Computer VisionTrackingReal-timeEdge

Smart Shelf Vision: Product Recognition on Edge

High-throughput recognition pipeline for retail shelf images with optimized inference and robust logging.

PyTorchOptimizationDockerLinux

Contact

Telegram

Open to full-time roles (EU / Remote).