About Me

Hello! My name is Maxwell Cai. My Chinese name is 蔡栩. I am currently a researcher and engineer in the Accelerated Computing Systems and Graphics Group (AXG) at Intel. With my background in computational astrophysics (Ph.D. 2016), my research interests include:

  • Develop physics-informed deep neural networks for tackling computationally-expensive (astro-)physical problems
  • Develop highly-scalable deep neural networks that can be trained on tens of thousands of processors and ingest TB-scale datasets
  • Research reinforcement learning, Bayesian inference, probabilistic programming, and computer vision, and apply them to astrophysical dynamics problems and medical imaging problems
  • Research explainable AI (XAI)

In a nutshell, I am particularly interested in non-linear dynamics and have been very keen to accelerate these calculations using scalable and interpretable deep neural networks. I believe that deep learning can be a powerful novel tool for scientific research if properly applied.

Prior to joining Intel, I was an advisor and researcher on high performance machine at SURF (Dutch National Supercomputing Center). Before stepping into the field of AI for Science, I obtained a Ph.D. in computational astrophysics from the Chinese Academy of Sciences and subsequently continued as a postdoctoral fellow at Leiden Observatory, Leiden University, The Netherlands. I have been researching the formation of planets and their long-term evolution to better understand the formation of our Solar System and many other solar systems orbiting other stars. Why is our Solar System the way it is? How was it form? How unique is our Solar System? Can we find other solar systems that support (intelligent) lives? What are the future of our Solar System and others? These are among the outstanding questions surrounding me as an astrophysicist. Unlike many other astrophysicists who observe the Universe through telescopes, I use (super-)computers as my primary research tool. According to the observational astrophysicist's observational data, I model these systems in the computer and carry out simulations. The simulation results can then be compared to the observational results and make predictions. As of July 2021, I have more than 20 peer-reviewed scientific publications in this field. I also co-authored a graduate-level textbook "Moving Planets Around" that teaches students how to build state-of-the-art N-body simulations from scratch and carry out research in this field. The book has been published by MIT Press in September 2020.

In my spare time, I appreciate Nature and enjoy road-trip traveling. I am also a big fan of classical music, especially the work from J.S. Bach.