AI for Physics and Optimization

Current Research

I lead research efforts developing statistical methods for generative modeling and optimization in the context of multiphysics and design. Currently, I am leading research efforts with the NanoML team at Purdue University and the AI for Quantum team at Quantinuum in developing latent polynomial and graph learning, machine learning assisted generative optimization models, and fundamental generative modeling techniques for multimodal graph network and multiphysics applications. Our research is highly interdisciplinary, and I am always looking for new collaborators. If you are interested in working with us, please reach out!

Generative Modeling for Device Design

Device design is a highly multidisciplinary field requiring knowledge all the way from applied mathematics and fundamental physics to algorithm design and programming to material fabrication and characterization to multiphysics and engineering. Improving device design requires optimizing the design process at any of these steps, from the initial concept to the final fabrication. When designing devices, we try to generate designs to optimize some objective function, such as computation time, power draw, spectrum, etc. In the past, solving design problems relied on the designers intuition, searching large design spaces or even sampling NP-Hard problems. To reduce the design problem's difficulty, we use latent models, such as LDMs, GANs, and VAEs, to extract symmetries and features in the data to construct a more efficient search space for design optimizers.

Generative Optimization and ML for Device Design

When designing devices, we try to generate designs to optimize some objective function, such as computation time, power draw, spectrum, etc. In the past, solving design problems relied on the designers intuition, searching large design spaces or even sampling NP-Hard problems. To reduce the design problem's difficulty, we use latent models, such as LDMs, GANs, and VAEs, to extract symmetries and features in the data to construct a more efficient search space for design optimizers.

Physics-Inspired Generative Modeling

Generative modeling constructs probabilistic models often by exploiting symmetries, constraints, and physics in data to reduce what needs to be learned.

Graph Learning

Graph networks are prevalent throughout mathematics, computer science, and engineering. Message passing, graph neural networks, and more recently graph transformers allow for the manipulation of graphs in machine learning.

Announcements

December 2024: PearSAN

Back in 2022, Michael Bezick joined our research group to work on machine learning.

November 2024: Congrats to Vea :)

Back in 2021, just as things were beginning to wind down from Covid, Sasha asked the group if anyone would be willing to chat with a potential new undergraduate researcher. I was just starting my third year but I knew I wanted to start a research lab to work on ML for quantum devices. Luckily, this bright new researcher was extremely optimistic about quantum. I decided to meet with her and it was evident she was beyond passionate about quantum technologies. I had never mentored an undergraduate student one-on-one before, let alone led research efforts (with Sasha's guidance of course). Congratulations to Vea for being accepted to the ECE PhD Program at Purdue :).

September 2024: Off to Cambridge

I traveled to Cambridge for Quantinuum's internal conference. Met some great scientists, including Adam Ollanik and the Colorado photonics team.

August 2024: RAPTOR

RAPTOR was published in Advanced Photonics. I was surprised, but grateful, of all the media coverage, including this one from Purdue with a great title. https://stories.prf.org/raptor-takes-a-bite-out-of-global-counterfeit-semiconductor-market/