ML Specialist / Developer · Artificial Intelligence Laboratory · Dubna State University
Computer Science & Engineering · ACM Certified Reviewer
research question → architecture → agents → experiments → evidence → publication
365 days of building, experimenting and shipping.
I build interpretable and trustworthy AI systems, combining model development, experimental validation and research software engineering.
My main interests are Explainable AI, neuro-fuzzy systems, structural interpretability, robustness, reproducibility and AI evaluation.
flowchart LR
R["AI Research"]
X["Explainable &<br/>Trustworthy AI"]
N["Neuro-Fuzzy &<br/>Interpretable Models"]
V["Evaluation &<br/>Robustness"]
E["Research<br/>Engineering"]
R --> X
R --> N
R --> V
R --> E
style R fill:#111827,stroke:#00F5FF,stroke-width:2px
style X fill:#111827,stroke:#FF2BD6
style N fill:#111827,stroke:#7B2CFF
style V fill:#111827,stroke:#00F5FF
style E fill:#111827,stroke:#FF2BD6
|
Trust Assessment through Dynamic Explainability Quantitative assessment of AI systems through explanation quality, robustness, concept drift and bias shift.
|
Hybrid GD-ANFIS / SHAP architecture integrating explainability with model training and validation.
|
|
Differentiable fuzzy attention for multimodal learning with interpretable reasoning mechanisms.
|
Deep neuro-fuzzy models and Routed Kolmogorov-Arnold Fuzzy Networks with structural interpretability and stability analysis.
|
|
Hybrid explainable architectures combining Kolmogorov-Arnold Networks, fuzzy systems and interpretable decision mechanisms.
|
Scientific vision pipeline for seismic fault segmentation using anisotropic geometry, topology and graph reasoning.
|
More research systems
Budgeted counter-evidence auditing for black-box time-series models with adaptive evaluation budgets.
Reproducible lifecycle management and evaluation of explainable AI models.
AutoXAI Reproducibility Model lifecycle
Desktop research environment for ANFIS analysis, training and visualization.
ANFIS Desktop ML Research tooling
Verified Explainability Core: A GD–ANFIS/SHAP Hybrid Architecture for XAI 2.0
Automatic Documentation and Mathematical Linguistics, 59(S5), S469–S478.
Hybrid and Hierarchical Explainable AI Based on Kolmogorov–Arnold Networks
Lecture Notes in Networks and Systems, Vol. 1763, Springer, 36–43.
DOI ↗
SynergiXAI Platform Architectural Model for Reproducible Lifecycle Management of Artificial Intelligence Models
Soft Measurements and Computing, No. 3-2, 101–119.
A Fuzzy Transformer for Multimodal AI: Differentiable Fuzzy Attention and Adaptive Explanations
Soft Measurements and Computing, No. 11, 34–55.
DOI ↗
HYBRID-XIRIS: Neuro-Fuzzy Architecture for Explainable Biometric Identification by Iris
Soft Measurements and Computing, No. 12-2, 119–132.
Routed Kolmogorov-Arnold Fuzzy Networks for Stable Interpretable Prediction
IITI'26 · Springer Proceedings
Concept-Latent Space Alignment in Fuzzy Attention Networks for Safety-Critical Intelligent Decision Support
IITI'26 · Springer Proceedings
AI coding agents are a primary implementation interface in my workflow.
I define the problem, system architecture, constraints and evaluation protocol. Agents accelerate implementation and experimentation; results are accepted only after testing and evidence-based validation.
flowchart LR
Q["Research<br/>Question"]
S["Architecture &<br/>Specification"]
A["Agent<br/>Orchestration"]
C["Implementation"]
E["Experiments"]
T["Tests"]
M["Metrics &<br/>Evidence"]
H["Human<br/>Validation"]
R["Research<br/>Artifact"]
Q --> S
S --> A
A --> C
A --> E
C --> T
E --> M
T --> H
M --> H
H -->|"refine"| S
H -->|"accept"| R
style A fill:#111827,stroke:#FF2BD6,stroke-width:2px
style H fill:#111827,stroke:#00F5FF,stroke-width:2px
style Q fill:#111827,stroke:#7B2CFF
style R fill:#111827,stroke:#00F5FF
Agents accelerate implementation. Evidence decides what survives.
|
|
|
|