Research

I study how baryonic physics affects weak gravitational lensing observables in cosmological simulations. My work combines hydrodynamical simulations, non-Gaussian lensing statistics, baryon correction methods, and reduced-variance emulation to prepare for LSST, Euclid, and Roman.

NSF Graduate Research Fellow · AAS National Osterbrock Leadership Fellow · Columbia CTL Lead Teaching Fellow


Research Themes

Weak Lensing Statistics

Power spectra, peak counts, and other non-Gaussian probes of the matter distribution. Quantifying how higher-order statistics improve cosmological constraints beyond two-point functions.

Baryonic Effects

Connecting hydrodynamical simulations to baryon correction models. Understanding how feedback processes — AGN, stellar winds, gas cooling — reshape the matter field and bias lensing observables.

Simulation & Emulation Methods

Reduced-variance techniques (CARPool), Gaussian-process emulators, and ML-driven tools for building fast, accurate predictions from expensive simulation suites like CAMELS-TNG.

Field-Level Modeling (BIND)

Developing BIND, a neural network-based field-level baryonification and intrinsic alignment model parameterized directly by IllustrisTNG subgrid and cosmological parameters — jointly correcting for feedback and IA in weak lensing analyses.

Selected Work

The impact of baryons on weak lensing statistics as a function of halo mass and radius
M. E. Lee, Z. Haiman, S. Genel · arXiv:2603.11815, 2026
Quantifies how halo mass and radius drive baryonic impacts on weak lensing statistics.
The effect of intrinsic alignments on weak lensing statistics in hydrodynamical simulations
M. E. Lee, Z. Haiman, S. Pandey, S. Genel · ApJ, 2026
Measures how intrinsic alignments contaminate non-Gaussian lensing probes in full-physics simulations.
Zooming by in the CARPool(GP) lane: new CAMELS-TNG simulations of zoomed-in massive halos
M. E. Lee, S. Genel, B. D. Wandelt, et al. · ApJ, 2024
Introduces variance-reduced Gaussian-process emulation of zoomed-in hydrodynamical simulations.
Comparing weak lensing peak counts in baryonic correction models to hydrodynamical simulations
M. E. Lee, T. Lu, Z. Haiman, J. Liu, K. Osato · MNRAS, 2022
Tests whether baryon correction models can reproduce non-Gaussian lensing statistics from full simulations.

All publications →


Software

CARPoolGP Gaussian-process regression with control-variates variance reduction for simulation-based inference.
BCM_lensing Baryon correction model pipeline for weak lensing peak count analyses.
cosmoANP Attentive neural process for fast cosmological field-level inference.
hydro_replace Tools for replacing dark-matter-only fields with hydrodynamical predictions.
zoomGZ Pipeline for generating and managing zoom-in simulation suites across high-dimensional parameter spaces.

GitHub →


Background

Before graduate school, my path was nonlinear: I spent years traveling, working on organic farms, and managing environmental non-profit offices before returning to Cabrillo Community College and then transferring to UC Berkeley (B.A. in Physics and Astronomy, 2020). I am now a PhD candidate at Columbia University, advised by Zoltán Haiman and Shy Genel.