Umer Gupta

London, UK

umer.gupta152@gmail.com

I work on geometric deep learning - generative geometric modelling, symmetry-respecting architectures, and topological priors.

My current research, with the University of Leipzig and ETH Zurich, develops autoregressive and diffusion-based methods for branching biological morphologies (botanical trees, neurons) under topological guidance. The first of this line of work appears at the GRaM Workshop at ICLR 2026; extensions to neuronal morphologies are in progress.

More broadly, I am drawn to deep learning applications in biomedicine, with a particular pull toward problems where geometric, structural, and dynamical priors matter. My areas of interest include protein structure and design, molecular generation and drug design, single-cell biology and perturbation modelling, and connectomics.

I hold an MSc in Data Science from the University of Edinburgh (Distinction, 2023) and a BSc (Hons) in Mathematics from Sri Venkateswara College, Delhi University. I am currently applying for PhD positions in the methodological and applied directions described above.

Alongside research, I lead ML at New Gradient, working on geospatial foundation models for ecosystem monitoring and subsurface modelling.

news

Apr 26, 2026	Presenting Autoregressive Frontier Expansion as a poster at the GRaM Workshop, ICLR 2026, in Rio de Janeiro.
Mar 02, 2026	Autoregressive Frontier Expansion: Growing Trees with Graph Machine Learning accepted to the GRaM Workshop at ICLR 2026.
Sep 10, 2025	Origin Peptides’ and New Gradient’s selected for Innovate UK’s £6.4M SMMIP programme - applying machine learning to inform real-time experimental optimisation for protein synthesis.
Sep 08, 2025	Started a research collaboration with ETH Zurich on generative models for branching biological morphologies.
Sep 01, 2024	Received an Innovate UK Nature-Positive AI grant for geospatial monitoring of peatlands at New Gradient.

selected publications

ICLR-W
Autoregressive Frontier Expansion: Growing Trees with Graph Machine Learning

Umer Gupta, Saku Peltonen, and Martin Ritzert

In GRaM Workshop, International Conference on Learning Representations (ICLR), Rio de Janeiro, Brazil, Apr 2026

Abs Bib

Tree-like branching structures are common in nature, from botanical trees to neurons and respiratory trees. Their branching shape often reflects function, making structural modeling central to understanding how these systems work. Acquiring real-world 3D data via imaging can be expensive or infeasible, so realistic generative models are valuable for simulation and augmentation. Existing approaches either rely on hand-tuned, mechanistic procedures, or do not jointly generate both the tree topology and its 3D geometry. We propose a graph neural network architecture that generates trees through an iterative expansion process, simulating the biological growth of real trees. At each step, it grows the frontier by predicting whether each active branch should bifurcate or terminate. Experiments on botanical trees show that our method can learn the 3D branching structure across multiple trees.
@inproceedings{gupta2026autoregressive, title = {Autoregressive Frontier Expansion: Growing Trees with Graph Machine Learning}, author = {Gupta, Umer and Peltonen, Saku and Ritzert, Martin}, booktitle = {GRaM Workshop, International Conference on Learning Representations (ICLR)}, year = {2026}, month = apr, location = {Rio de Janeiro, Brazil} }