Aleksander Mądry

Mądry helped establish adversarial training, which teaches AI systems to remain reliable when inputs are deliberately altered to fool them. His MIT group has made robustness and safe real-world deployment central problems in modern machine learning.