Saed Rezayi

NLP research scientist · trustworthy evaluation of language models

I’m an NLP research scientist working on language models in high-stakes settings, currently at NBME.

My research starts from a simple idea: almost everything we believe about what AI systems can do rests on a measurement. I study whether those measurements are telling us the truth, and how to rebuild them when they are not. This matters especially when AI systems evaluate other AI systems, because a failure may come from the system being tested or from the way we measured it.

Research Judge validity, explainability, fairness, robustness, item generation Publications 20+ papers spanning model evaluation, fairness, explainability, and knowledge-infused NLP About Background, and how I ended up working on this Photography Street work, mostly at the wrong shutter speed