Trustworthy machine learning
Reliable AI systems whose behavior can be evaluated beyond benchmark accuracy.
Research profile
Research, projects, and grants
I develop AI systems that remain dependable when data shift, unknown categories appear, inputs are corrupted or adversarial, and models must continue learning.
Research vision
I am an Associate Professor at New Uzbekistan University and a Visiting Researcher at the Okinawa Institute of Science and Technology (OIST), Japan. My research focuses on trustworthy machine learning, anomaly detection, out-of-distribution detection, continual learning, and AI robustness. Previously, I led AI research initiatives at the Institute for Research in Fundamental Sciences (IPM) in Tehran, Iran, and worked as a Senior Researcher in Finland and France, including at the Center for Machine Vision and Signal Analysis (CMVS). My work bridges machine learning foundations and practical robustness for reliable, secure, and effective AI systems.
My work connects anomaly and out-of-distribution detection, robustness, continual learning, and AI security to move trustworthy machine learning from accuracy alone toward reliable behavior in the real world.
Research interests
The four themes connecting my theoretical work, evaluation methods, and practical AI systems.
Anomaly and OOD detection
Recognizing unfamiliar, rare, shifted, or abnormal inputs.
Continual and lifelong learning
Learning over time while preserving useful knowledge and managing change.
Robustness, safety, and reliability
Studying adversarial, corrupted, backdoored, and failure-prone AI behavior.
Current workplaces
Research and collaboration connecting Uzbekistan and Japan.
New Uzbekistan UniversityAssociate Professor · Tashkent, Uzbekistan
Okinawa Institute of Science and TechnologyVisiting Researcher · Okinawa, JapanCampus photos: New Uzbekistan University · OIST / CC BY 2.0
Featured research
Selected work on dependable perception and recognizing unknown conditions.
CRANE
Context-guided prompt learning and attention refinement for zero-shot anomaly detection.
FrameShield
Adversarially robust video anomaly detection for more dependable visual monitoring.
Universal Novelty Detection
Adaptive contrastive learning for detecting novel data across datasets and tasks.
Research projects and grants
Japanese national research funding and institutional research programs.
- Investigating the Trustworthiness of Deep Pre-trained and Self-Supervised Models
2024–2028 · ¥4,680,000 · Principal Investigator
Japanese national research funding: JSPS KAKENHI, Grant-in-Aid for Early-Career Scientists · Funded at OIST, Okinawa, Japan · Official JSPS KAKENHI program - Breaking Boundaries: Robust, Domain-General Anomaly Detection with Vision-Language Models
2026–2030 · ¥18,330,000 · Principal Investigator
Japanese national research funding: JSPS KAKENHI, Grant-in-Aid for Scientific Research (B) · Funded at OIST, Okinawa, Japan · Official KAKEN record (26K02988) - AI Safety and Security for Classical AI Models
Institute for Research in Fundamental Sciences (IPM)