Keming Zhang

I’m an Eric and Wendy Schmidt AI in Science Postdoc Fellow at the University of California, San Diego. I acquired my PhD in Astrophysics from the University of California, Berkeley and was previously an undergraduate at Columbia University.

My research primarily focuses on the intersection between AI and Astronomy, which includes gravitational microlensing, exoplanets, and simulation-based inference (SBI). I use deep neural networks to build surrogate inverse models that enables massively scalable inference required by modern-day astronomical survey experiments. My open-source software project nbi integrates importance sampling with Neural Posterior Estimation to enable asymptotically exact inference (AEI), and can be applied off the shelf for a wide range of astronomical inverse problems.

In my past work, I have applied AI to drive theoretical discovery in astrophysics, as highlighted in the news article “AI providing new light on lensing degeneracies”. By adapting SBI for analyzing large volumes of microlensing simulations, I have discovered a new physical degeneracy that is ubiquitous in interpreting planetary microlensing events. In two subsequent papers, I demonstrated how this new degeneracy revealed fundamental symmetries in two-body microlensing, which then led to discovery of analytic solutions for this problem that would allow for substantially faster forward models.

I’m also an observer and I use the Keck observatory to characterize microlensing planets. Apart from this, I have also been an astrophotographer since childhood. Check out my portfolio.