Postdoctoral researcher @ Caltech
nzilberstein@rice.edu · Google Scholar · Git · X · LinkedIn
I am a postdoctoral researcher at Caltech, working with Katie Bouman.
My research is on generative modeling, inverse problems, and sampling, mostly for imaging. Broadly, it falls into two categories: inference-time methods, where I develop sampling and optimization algorithms that use large generative models as priors to steer generation, and training new models tailored to a downstream task. Beyond images, I have applied these ideas to graphs and wireless systems.
I received my PhD and M.S. in Electrical and Computer Engineering from Rice University, where I was advised by Santiago Segarra. During my PhD, I interned at Google (Computational Imaging team, 2026), the Center for Computational Neuroscience at Flatiron Institute (LCV group, 2025), and InterDigital (2024). Before Rice, I studied Electrical Engineering at the Universidad de Buenos Aires, in my hometown of Buenos Aires, Argentina.
News
- Sep 2026 — Paper (Flow map denoisers) accepted to NeurIPS 2026
- Sep 2026 — I successfully defended my PhD
- May 2026 — I started my internship at Google, working with the Computational Imaging team
- Feb 2026 — Paper (Energy models for inverse problems) accepted at ICML 2026. This paper is the outcome of my summer internship at Flatiron!
- Dec 2025 — I am honored to receive the NVIDIA Academic Grant as a co-author to support our work entitled “Test-time Scaling with Ensemble Kalman for Inference via Diffusion Models”
Selected publications
Normalized Energy Models for Linear Inverse Problems —
Flow Map Denoisers: Traversing the Distortion-Perception Plane for Inverse Problems —
Repulsive Latent Score Distillation for Solving Inverse Problems —