Developer Behavior Modeling & Personalization
Led behavioral intelligence research for AI-assisted software engineering, developing methods to infer developer intent, AI usage patterns, workflow habits, and recurring task routines from repository and interaction evidence. This work resulted in two research papers.
Adaptive LLM Code Assistance
Led the design and evaluation of context-aware developer assistance spanning adaptive invocation, suggestion timing, output filtering, contextual support, and multi-model LLM serving for code completion, chat, and repository-aware assistance. This work contributed to two research papers.
ITMO Code — AI Education Platform
Transformed the open-source Roo-Code coding extension into ITMO Code, a full-stack AI development platform for education. Designed and implemented editor-native AI assistance, role-based workflow orchestration, Git/GitHub automation, sandboxed execution, agentic code review, backend APIs, a database layer, workers, analytics, and automated pipelines.
Agentic Systems & Evaluation
Conducted research on repository-grounded coding agents, issue localization, repository-level software repair, agent evaluation, and stage-aware pipelines. Evaluated how model effort can be allocated across exploration, localization, repair, and validation, connecting directly to my PhD direction in Adaptive Human-AI Collaboration via Agentic Systems.
Multimodal Data & Web Agents
Developed agentic systems for conversational data analysis and browser automation, including sandboxed execution, agent orchestration, multi-step reasoning, visualization, and multimodal interaction.
Research dataset scale
Repository and developer-interaction records used across my research on developer behavior, personalization, and AI-assisted software engineering. These figures summarize the coverage of the research datasets.