Nicholas Geneva

Timeline

1. 2015

   Lattice Boltzmann at Delaware

   In November 2015, he and Lian-Ping Wang, both at the University of Delaware, are the authors of an oral talk at the 68th Annual Meeting of the APS Division of Fluid Dynamics. The talk covers algorithms for the collision and streaming steps of lattice Boltzmann simulations, with performance compared on a 3D turbulent channel flow.

2. 2019

   Physics-constrained auto-regressive networks

   In June 2019, he and Nicholas Zabaras post a paper on modeling PDE dynamics with physics-constrained deep auto-regressive networks. The model is an auto-regressive dense encoder-decoder convolutional network with a Bayesian framework for uncertainty, trained without training data, and the paper is accepted to the Journal of Computational Physics.

3. 2020

   Generative models of turbulent flows

   In December 2020, he and Zabaras publish "Multi-fidelity generative deep learning turbulent flows" from the Scientific Computing and Artificial Intelligence (SCAI) Laboratory at the University of Notre Dame. The model is a conditional invertible neural network with recurrent LSTM connections that generates turbulent flow fields from an inexpensive low-fidelity solution.

4. 2020–2022

   Transformers for physical systems

   At Notre Dame's SCAI Laboratory, he and Zabaras post "Transformers for Modeling Physical Systems" in October 2020, and it appears in Neural Networks in February 2022. Koopman-based embeddings project a dynamical system into a vector representation that a transformer predicts, and the model outperforms classical methods common in scientific machine learning.

5. 2023

   ClimSim's Outstanding Paper award

   In 2023, ClimSim, a dataset for hybrid physics-ML climate emulation that he coauthors, receives a NeurIPS Outstanding Paper award. The dataset holds 5.7 billion pairs of input and output vectors from multi-scale climate simulations developed by a consortium of climate scientists and ML researchers.

6. 2024

   Earth2Studio for AI weather

   In 2024, he and Dallas Foster are credited as the authors of NVIDIA Earth2Studio, an open-source deep-learning framework for exploring, building and deploying AI weather and climate workflows.

7. January 2026

   The StormScope paper

   In January 2026, he is one of 13 authors of "Learning Accurate Storm-Scale Evolution from Observations". The paper introduces Stormscope, a family of transformer-based generative diffusion models trained on geostationary satellite imagery and ground-based weather radar over the continental United States, with forecasts at 10-minute and 6-kilometer resolution for lead times up to 6 hours.

8. As of October 2026

   Senior engineer on Earth-2

   He is a Senior Software Engineer in HPC/AI at NVIDIA, where he develops software for Earth-2 that supports AI-driven weather and climate prediction. He has been a core developer of NVIDIA's PhysicsNeMo Python packages for four years, and his SciPy 2026 talk covers software orchestration and pipelines for AI weather forecasting with Earth2Studio and PhysicsNeMo.