Mikhail Zemskov

Automated ingestion and preparation of terabyte-scale training data for 3D object detection, developed and maintained processing pipelines and quality checks. Investigated 3D detection approaches, integrated models into a real-time system, and improved the real-time detector's mAP by 2% relative to baseline.

Developed and deployed semantic-search models for e-commerce, from data preparation and synthetic example generation to training and validation pipelines. Built embedding models that delivered an average 45% relative improvement in NDCG over baseline across 3 clients.

Developed and maintained 10+ ETL pipelines and Data Vault data marts for financial modeling, processing terabyte-scale datasets. Contributed to enterprise data warehouse architecture, prepared code and artifacts for releases, and supported pipelines and data marts in production.

Built license plate recognition for challenging imaging conditions and OCR for Japanese documents; developed the Strelok-CV automated judging system and video emotion recognition.

Independently designed and implemented ETL pipelines for regulatory reporting on terabyte-scale data, covering source analysis, Spark ingestion from PostgreSQL into HDFS, and output validation. Prepared data extracts and Qlik Sense reports; checked and analyzed the resulting datasets.

Developed and maintained 10+ streaming and batch pipelines for telemetry, user activity, and structured data at terabyte scale. Built data delivery and quality-checking procedures and contributed to enterprise data management infrastructure.

Automated ETL workflows for ML training data and developed a Data Vault warehouse; implemented testing and data quality checks, standardized development workflows, and contributed to MLOps practices.

Built MVP components for cloud applications on an open-source stack and performed integration testing using OpenStack, Docker, Podman, Java, and Open Liberty.

Contributed to an EEG-based system for early detection of depressive disorders: analyzed data and research literature, implemented models, and built preprocessing for data from multiple sources using Python and PyTorch.

Specialist Degree in Integrated Aircraft Systems from Moscow Aviation Institute (MAI) from 2018 to 2024.