I lead AI architecture and R&D at the intersection of production machine learning, edge intelligence, and robotics. At KPI, I research robotic perception, visual navigation, and autonomous UAV systems; at Samba TV, I lead AI system design across cloud, edge, and data workflows. My work focuses on turning computer vision research into reliable systems for video understanding, autonomous perception, and media-intelligence products.
Email · Publications · Google Scholar · LinkedIn
Russia's war against Ukraine is ongoing. Support Ukraine · war.ukraine.ua
Current work
- AI architecture and technical leadership for production ML systems.
- Edge AI and video intelligence: model deployment, optimization, evaluation, and MLOps.
- Robotic perception, visual navigation, and coordination for autonomous UAV systems.
Selected work
- Depth-Assisted Sparse Visual Odometry for UAV-Relevant Synthetic RGB-D Evaluation. Robotics, 2026. paper · code · dataset
- Edge Intelligence Resource Consumption by UAV-based IR Object Detection. UAVM, 2023. paper
- Object Detection for Rescue Operations by High-Altitude Infrared Thermal Imaging Collected by UAVs. IbPRIA, 2023. paper
Recent writing
- Depth-Assisted Sparse Visual Odometry: A Controlled Geometric-Backend AblationAnnouncement of my new Robotics paper on depth-assisted sparse visual odometry.
- Efficiency in AI: Practical tips for optimizing AISlides from my public talk on practical AI efficiency.
- How Object Detection Evolved: From Region Proposals and Haar Cascades to Zero-Shot TechniquesSlides and recording from my talk on the evolution of object detection.
- ChatGPT in Machine LearningAn early look at how ChatGPT explains machine learning concepts and where to be cautious.
- Classification Metrics: Visual ExplanationsVisual explanations of accuracy, precision, recall, F1, ROC curves, and AUC.
- Lightweight Neural Network ArchitecturesSlides and recording from my talk on lightweight neural network architectures.
- Welcome!A short note on why I started writing about my machine learning work.