Keming Zhang

I am a NASA Hubble Fellowship Program (NHFP) Sagan Fellow at MIT. In Fall 2026, I will move to the Institute for Advanced Study in Princeton as a Member, where I will continue as an NHFP Sagan Fellow. Previously, I was an Eric & Wendy Schmidt AI in Science Postdoctoral Fellow at UC San Diego (2023–2025). I received my Ph.D. in Astrophysics from UC Berkeley in 2023 and my B.A. in Astrophysics from Columbia University in 2018.

My research aims to build a complete statistical picture of cold exoplanets and their evolution over time: from free-floating planets ejected from their birth systems, to planets that outlive their dying hosts. To this end, I use gravitational microlensing as my primary tool, combining observational, analytical, and statistical approaches.

Most microlensing planets have unknown physical masses and host properties. I lead a long-term Keck Adaptive-Optics observing program to characterize microlensing planet hosts, which has been awarded 12 half-nights over four years (2023–2026). A highlight from this program includes the discovery of an Earth-sized exoplanet orbiting a white dwarf at 2.1 au, “suggesting a future in which our planet outlives its star” (New York Times). We are also conducting a multi-epoch survey to test whether free-floating planet candidates found by microlensing may be bound planets on very wide orbits (≳20 au) instead. More recently, I identified a sharp truncation near mass ratio 2% in the cold exoplanet mass-ratio function, providing evidence for scale-invariant giant-planet formation via core accretion.

I extensively integrate AI into my research — from developing custom neural networks to working with LLM agents. By analyzing microlensing simulations using an AI inference framework, I discovered a unified theory of planetary microlensing degeneracies — that most microlensing planets are subject to a two-fold degeneracy in projected separation — which I subsequently proved analytically. Currently, I am working with LLM agents to prove my numerically verified conjecture that planetary microlensing is universally approximated by a specific variable-shear Chang-Refsdal lens, which admits closed-form solutions.

Outside of research, I have been an avid astrophotographer since childhood. Check out my portfolio.