Trustworthy and uncertainty-aware ML
Calibration, uncertainty estimation, failure analysis, and robust evaluation for high-stakes prediction tasks.
Alexei Kornaev
Research Scientist in Physics and AI, including Deep Learning, Computer Vision, Physics-informed neural networks, AI in Medicine, and Robotics
Alexei Kornaev works on machine learning and computer vision for scientific and clinical settings, with a focus on medical imaging, radiogenomics, trustworthy AI, physics-informed learning, and robotics. His research combines method development with questions of uncertainty, robustness, interpretability, and deployment in high-stakes environments.
Research agenda
The central theme of this work is reliable machine learning where physical constraints, medical context, and real-world variability matter. This includes uncertainty-aware prediction, physics-informed modeling, multimodal analysis of imaging and molecular data, and perception-action pipelines for embodied systems.
Physics-informed learning
Neural and variational methods for scientific computing, inverse problems, and flow modeling in complex physical systems.
Medical AI and radiogenomics
Joint analysis of medical imaging, pathology, molecular information, and clinical context for diagnosis and prognosis.
Robotics and embodied AI
Perception, planning, and vision-language-action systems for robotic assistance and interactive environments.
Selected directions
Medical imaging and multimodal diagnosis
Deep learning for radiology, pathology, and related multimodal pipelines that connect image evidence with clinical and biological context.
PINNs and scientific machine learning
Physics-informed neural networks, neural surrogates, and variational formulations for computational mechanics and fluid-flow problems.
Reliable AI in practice
Generalization under shift, uncertainty quantification, reproducibility, and evaluation protocols for deployment-oriented ML.
Vision, language, and action
Multimodal systems that connect perception with reasoning and control in robotics and interactive AI settings.
Current affiliations
- Center for Top-Level Educational Programs in AI, Innopolis University, Innopolis, Republic of Tatarstan, Russia
- N.N. Blokhin National Medical Research Center of Oncology, Moscow, Russia