A resource-aware experimental framework for selecting, fine-tuning, profiling, compressing, and ranking computer vision models for edge deployment. The current milestone focuses on YAML-driven classification experiments for AID aerial scene data, torchvision model candidates, FP32 fine-tuning, checkpoint profiling, result traceability, hard-constraint filtering, Pareto analysis, and weighted ranking.
Portfolio signal: edge AI, computer vision, reproducible experimental protocols, deployment constraints, and decision support for choosing practical models rather than only the most accurate architecture.
A reproducible research stand for adaptive control of heterogeneous distributed computing systems. It includes a multi-agent model, load and failure scenarios, scheduling algorithms, ML forecasting, a ZNN balancing block, an LLM-agent policy loop, CLI, web UI, reproducibility protocols, and quality gates.
Portfolio signal: a direct connection between the scientific topic, software engineering, and experimental infrastructure.
A local research stand for migration text analysis and urban perception studies. It includes Telegram/YouTube CSV corpora, toponym and urban-space analysis, migration narratives, reproducible samples for manual coding, retrieval/summarization over scholarly articles, and publication discovery through open APIs.
Portfolio signal: an evidence-first research workflow where data, papers, experiments, and code live in one project environment.
A backend prototype for expert analysis of interactive publications. The system analyzes a page by URL, extracts features, applies an assessment rubric, produces deterministic reports without API keys, and optionally generates AI-assisted reports through OpenAI, Gemini, or Ollama.
Portfolio signal: an applied bridge between evaluation methodology, backend API design, and interchangeable AI providers.
A simulation of distributed computing system control with carbon-footprint reduction in mind. It includes synthetic data, large-network modeling, comparison of ZNN, LNN, and hybrid ZNN-LNN workflows, energy and emissions calculations, and result visualization.
R&D on computer vision for UAS (2023-2024) — preparation of proposals and technical specifications for autonomous navigation, multispectral optoelectronic systems, autonomous detection, and monitoring.
On-board CV (2025) — an experimental track for computer vision on UAVs: comparison of Jetson Nano and Radxa CM5, preparation of datasets and test logs, pre-flight preparation, and flight mission generation.
Geoinformation and coordinate-time support for the transport sector (2025) — a concept for a unified digital platform for spatial data, navigation, and intelligent transport infrastructure management.
Russian Railways: station scheme recognition and 3D modeling (2025) — design work for an information system with modules for recognizing railway station schemes and building 3D models of infrastructure objects.
Urban Environment, Roads, and Natural Risks
Automated GIS for residential complexes in Krasnoyarsk (2021) — a geoinformation system for assessing residential-complex comfort, urban environment quality, transport accessibility, infrastructure, and building parameters.
Intelligent Traffic Management System (2024-2025) — a software platform for supporting traffic management and data exchange between system modules; software registration certificates No. 2024610926 dated 16 Jan 2024 and No. 2025610587 dated 13 Jan 2025; demo.
Automated information and measurement system for road water-thermal regimes (2024) — a road-condition monitoring system using meteorological and soil data to forecast reductions in the bearing capacity of road structures.
Intelligent flood monitoring and forecasting system (2025) — technical specification and MVP estimation for a system that collects and analyzes open hydrological data using hybrid artificial intelligence models.
AI, Data, and Industrial Analytics
Video stream analytics system (2024) — software for classifying human actions in images and assessing potentially abnormal situations.
IntelliMDM (2025) — a concept for an intelligent master data management platform based on predictive analytics, LLM technologies, and centralized enterprise data storage.
Center for Digital Industrial Solutions (2025) — development of a shared-use center concept and infrastructure for industrial digital solutions, computing tasks, and GPU-based workflows.
AI-driven adaptive heating for sodium-ion batteries (2026) — an international R&D track with Tianjin University of Technology on intelligent management of sodium-ion batteries for truck starting power supplies in cold regions.
Cultural Heritage, AR/VR, and Humanities Projects
AR/VR for tourism and natural heritage (2021) — concepts for information systems that preserve and represent natural and cultural-religious heritage, including augmented reality scenarios.
Spiritual Ring of the Yenisei North (2026) — a concept for a documentary travel film about Orthodox heritage in the Yenisei North.
International collaboration with Thammasat University (2026) — preparation of formats for research and education cooperation, project exchange, and joint humanities research.