About Me

I’m Minoh Jeong, a postdoctoral researcher at the University of Michigan. Broadly, I study how to turn complex, high-dimensional, and often noisy data into reliable, useful representations. My work sits at the intersection of machine learning, information theory, and statistical signal processing, guided by a simple question: what makes a learned representation trustworthy and informative?

I approach research with equal parts theory and practice. I enjoy building clean formulations, proving what we can (and cannot) guarantee, and then testing those ideas on real data. Much of my recent focus is on robustness-handling distribution shift, label noise, and data heterogeneity and on evaluation, designing measurements that reflect what we actually care about. I’m especially drawn to multimodal problems where text, images, audio, and other signals must be aligned and reasoned about together, as well as applications in biostatistics and health where uncertainty truly matters.

I value collaboration and clarity. Many projects begin with a whiteboard, a few counterexamples, and a commitment to simple baselines before chasing complexity. I aim for solutions that are conceptually minimal yet empirically strong, and I care about transparent methodology, reproducibility, and careful ablations that separate signal from noise.

Looking ahead, my goal is to build a research program that advances principled representation learning and its applications, while fostering open, rigorous collaboration across disciplines. If your interests overlap or if you have a problem that could benefit from careful modeling and clear evaluation, I’d love to connect.

During my free time, I like doing activities. I usually go to gym and practice calisthenics.