Machine learning research scientist
PhD in computer science, Hebrew University of Jerusalem
I am interested in the underlying principles that give rise to biological complexity, and in how information emerges through machine learning paradigms. Across my academic and industry work, I have been fortunate to study these fundamental questions and to apply what I learned across fields, including video editing, autonomous driving, dynamical systems and cell foundation models.
Built agentic and ML pipelines that automatically edit recorded episodes.
Quantified confidence in transcriptomic foundation models.
Advised by Dr. Mor Nitzan. Machine learning for dynamical systems and single-cell biology. Council of Higher Education and CIDR PhD fellowships.
Reconstructed tissue structure from single-cell data with optimal transport (NovoSpaRc).
Built real-time deep learning detection and scene-understanding models for autonomous vehicles.
Advised by Prof. Nir Friedman. Adapted single-cell RNA-seq to yeast, both technologically and computationally. 2nd in class, with an exchange at the University of Toronto.
Experimented with single-cell RNA-seq methods and their analysis in yeast, and designed a CRISPR combinatorial perturbation library.
Quantified how much of the variance in yeast stress response is explained by cell-cycle variation. Magna cum laude, Dean’s Award.
* equal contribution. Full list on Google Scholar.
NeurIPS 2025 workshop AI for virtual cells and drug discovery (AI4D3)
Can a transcriptomic foundation model tell a real cell from a shuffled one? We propose a model-agnostic test based on synthetic samples and pretraining loss. Models separate real from shuffled cells and show loss barriers between distant cell types, but a coherent world model is far from universal.
ICLR 2024
A deep-learning framework that classifies dynamical regimes and detects bifurcation boundaries by learning topologically invariant features, from electrical circuits to gene expression in pancreatic development.
ICLR 2023 Oral
An unsupervised embedding of dynamical systems that classifies them into physical classes such as stability and energy conservation, and predicts equations from data, with an application to meteorological data.
ISMB 2024 Bioinformatics
With a fixed sequencing budget, should you sequence more cells or each cell more deeply? We show reconstruction accuracy scales with the logarithm of either, and that the best allocation follows a power law in the budget.
Cell Reports 2023
How do tissues organize division of labor in space? We contrast spatial gradients with cell–cell interactions and find signatures of both, in intestine enterocytes and colon fibroblasts.
Nature Biotechnology 2023
A framework that transfers annotations from reference data to single-cell and spatial omics at scale, supporting expression decomposition, cell fate prediction and recovery from read dropout. We used it to map spatially defined functional units in colorectal cancer (eLife 2025).
Nature Protocols 2021
An optimal transport algorithm that reconstructs tissue structure from single-cell RNA-seq, optionally guided by a reference atlas, with a step-by-step guide on the Drosophila embryo.
Hebrew University of Jerusalem
I have a precious little daughter, Naya.
Sports keep me sane: I enjoy running, swimming and hiking. I love experimenting with arts, including woodworking, pottery, glass art and printing, and with regrowing all kinds of plants.
These are my awesome brothers: postdoc Avraham Moriel, MD PhD Nadav Moriel, and (fresh scholar) Gilad Moriel.