Austin Kaiji Wang

Hi! I’m Austin, a Caltech CS graduate and researcher at ByteDance Seed, where I currently work on post-training for visual foundation models. During my undergrad, I was fortunate to do research in the labs of Prof. Yisong Yue from Caltech and Prof. Stefano Ermon from Stanford.

My research focuses on developing efficient and principled algorithms that unlock new capabilities from generative models. I am particularly interested in post-training, inference-time scaling, and controllable generation: how to adapt powerful models to better satisfy rewards, constraints, and downstream objectives. More broadly, I aim to design methods that make generative models more capable and accessible across creative and scientific applications.