Direction

I am driven by the challenge of understanding the internal mechanisms of advanced machine learning models. While opaque optimization has achieved impressive capabilities, this success has come at the cost of understanding the algorithms these models actually learn. As these systems are increasingly deployed in high-stakes settings, this opacity becomes a critical concern. My career's north star is improving our understanding and control of advanced AI systems.

My technical machine learning research spans activation steering, adversarial robustness, machine unlearning, and pre-training data interventions. I'm particularly passionate about developing lightweight interventions that can address safety issues during deployment without requiring costly retraining - similar to how security patches provide critical safety benefits in traditional software. My research has been featured in leading venues including ICLR and TMLR. 

I'm currently seeking research opportunities, including full-time roles and fellowships. Don't hesitate to reach out!