What makes AI and ML so powerful? What is "large" about Large Language Models? Big optimizations, which contain at least tens of billions of parameters! In our group, we do "big" optimizations for physics and engineering problems, routinely dealing with at least hundreds of thousands of parameters to billions (and approaching the trillion mark one of these days). This is the concept of scientific inverse design: backpropagating through the laws of physics (e.g., Maxwell equations) and discovering new devices from scratch, whose "emergent" geometries and abilities challenge the status quo. We thrive at the nexus of wave physics, information processing, and high-performance computation with a wide range of interests, encompassing optics and photonics, light-matter interactions in classical and quantum environments, acoustics and phononics, bio-chemical reaction-diffusion systems, machine learning, computer vision and robotics (to name just a few)—in short, we care about any design problem which could benefit from rich physical equations and from carefully-formulated optimization and control. We are an extremely forward-looking group with a lot of new ideas and projects. As much as we celebrate the work we have done, we even more eagerly look forward to the work we will do, and relish the next challenge. Contact us for available positions (PhD and post-doc openings).
Check out a few highlights from our research (more slides on the way!).
While our ambitions are geared towards realizing novel systems and devices, our spirits revel in equations and programming. If you want to dig a little deeper into our techniques, please check these out (coming soon!)