Tianyuan Cai

I am a machine learning researcher studying how AI systems can learn from human behavior at scale and adapt safely to evolving human needs. My work explores how to build models that respect diverse human perceptions, offer individuals meaningful control, and remain predictably useful over time. I first began studying large-scale human behavior through economics research at Pomona College, later scaling these methodologies through research and product roles at Adobe Research, Firefly, and OpenAI. Today, my approach bridges the gap between machine learning and human systems, combining post-training and safety research, rigorous evaluation, and human studies at the scale of human diversity.