Published multiple papers at ICML, NeurIPS workshops on topics including AI red teaming, steering language model refusals, and systematizing security vulnerabilities in computer-use agents
Co-developed open-source tools: PyRIT (for generative AI red teaming) and DEQGAN (a GAN-based method for solving differential equations)
Paper 'Lessons From Red Teaming 100 Generative AI Products' presented at Microsoft BlueHat 2024 and NeurIPS 2024 Workshop
Contributed to Phi-3 safety post-training alignment using a 'Break-Fix' cycle
Published interdisciplinary work in AI for Science, differential privacy, urban sound classification, and humanitarian AI applications