projects
Engineering trustworthy AI systems — the agents, analyses, and verification tools behind my papers.
Trustworthy AI Agents for Code
Research on code agents that combine model reasoning with program evidence, verification, and developer oversight.
RefineAct
A runtime verification framework for checking an LLM agent's proposed actions before execution.
LocalizeAgent
An evidence-grounded agent for localizing design issues in Java programs using structured program facts and iterative reasoning.
MMAssist
An IntelliJ IDEA assistant that recommends Move Method refactorings and delegates behavior-preserving edits to the IDE engine.
Reliable & Verifiable AI
Research on analysis, repair, and modularity techniques for making AI systems measurable, maintainable, and verifiable.
NeuralState
A typestate analysis for statically detecting deep-learning API misuses before model training or execution.
IRepair
A targeted repair method that localizes and edits faulty LLM behavior while limiting disruption to general model performance.
RNN Decomposition
A modularization technique for decomposing trained RNNs into reusable and replaceable behavioral components.