Undergraduate at Halıcıoğlu Data Science Institute - UC SanDiego
Profile photo credit to Scott Yang at the Salk Institute, taken at SfN 2025.
Hi! My name is Kaiwen Bian and I am a 4th year undergraduate student at UC SanDiego double majoring in Data Science and Cognitive Behavoral Neuroscience.
Data in the right manifold is all you need: I believe that data is the fundamental bedrock of intelligence, but it is the manifold (the underlying low-dimensional geometry hidden within that data) that unlocks true generalization and robustness. My research is driven by the conviction that studying neuroscience is the most effective way to inform these AI systems; by deciphering how biological neural populations coordinate to extract structure from sensory chaos and react to them by interacting with the world, we can build agents that operate with the same efficiency and fluency. By sitting at the intersection of computational neuroscience, embodied AI, and representation learning. I focus on leveraging the manifold structure in neural and behavioral data to build more generalizable and robust AI systems.
Outside of my academic life, I enjoy skateboarding 🛹, surfing 🏄, and immersing myself in the serenity of the natural world 🌲
Developing computationally efficient deep RL imitation systems (Mimic-MJX) for bio-mechanically realistic agents, discovering topological structure in embodied behavior spaces (TopoMimic), and building neural-aligned embodied agents (Coming).
Creating latent dynamical models for planning and generating bio-mechanically realistic behaviors using deep state-space modeling methods. Co-advised on the neural-aligned embodied agent project from Salk.
Graph compositional abstraction for molecular generation (MOSAIC) and topological analysis of embodied agent behaviors. Co-advised on the TopoMimic project from Salk.