At HSE University, I’m researching neural runtime monitoring for the safety of LLM agents performing financial operations. I formalise safety invariants in LTL and check them with NuSMV and SPIN, while training a PyTorch monitor to predict violations.
In my research on financial fraud detection, I developed and compared machine learning models on transaction datasets and used SHAP to interpret predictions. I’ve also worked on empirical comparisons of LLM agent architectures and built projects involving distributed databases and software testing.
