Representation centered manufacturing intelligence

Modern manufacturing systems generate large volumes of sensor, machine, and simulation data, yet extracting knowledge is challenging. Our research develops interoperable representations that translate heterogeneous physical signals into compact descriptions. These representations provide a common interface for sensor fusion, virtual sensing, generative modeling, human-AI teaming, and agentic AI for monitoring, reasoning, and control. This vision motivates our research agenda on manufacturing systems integration through universal representation layers.

Representative work: Coupled Flow Matching · Diffusion in Aligned Tensor Space ·