StepShield
Research on AI agent safety — determining when, not whether, to intervene on rogue agents.
Co-authored research paper on AI agent safety, focusing on determining the optimal timing for intervention on rogue agents rather than whether to intervene at all.
Published as an arXiv preprint (arXiv:2601.22136) with collaborators from Stanford University, Indian Institute of Science, and University of the Cumberlands.
Tech Stack: Python, PyTorch, Machine Learning, AI Safety.