ML Engineer (converted from internship)
T-Bank
Russian commercial bank and financial services company.
Dec 2025 — PresentBuilt and productionized ML systems for tagging and catalog matching; the tagging pipeline achieved 0.95 recall and 0.98 accuracy and was projected to save about $15,000 in LLM API spend.
• Owned an end-to-end venue tagging pipeline from taxonomy design to deployment: defined high-level and low-level user-facing tags, trained review-based models to predict them, reached 0.95 recall and 0.98 accuracy, and saved about $15,000 in projected LLM API spend.